Earth and Planetary Science Letters 199 (2002) 389^402 www.elsevier.com/locate/epsl Moho topography in the central Andes and its geodynamic implications X. Yuan a; , S.V. Sobolev a , R. Kind a;b a GeoForschungsZentrum Potsdam, Telegrafenberg, 14473 Potsdam, Germany b Freie Universita«t Berlin, Berlin, Germany Received 26 June 2001; received in revised form 12 December 2001; accepted 4 March 2002 Abstract P-to-S converted waves at the continental Moho together with waves multiply reflected between the Earth’s surface and the Moho have been used to estimate the Moho depth and average crustal Vp /Vs variations in the central Andes. Our analysis confirms and significantly complements the Moho depth estimates previously obtained from wide-angle seismic studies and receiver functions. The resulting crustal thickness varies from about 35 km in the forearc region to more than 70 km beneath the plateau and thins (30 km) further to the east in the Chaco plains. Beneath the Andean plateau, the Moho is deeper in the north (Altiplano) and shallower in the south (Puna), where the plateau attains its maximum elevation. A non-linear relation exists between crustal thickness and elevation (and Bouguer gravity), suggesting that the crust shallower than 50^55 km is predominately felsic in contrast to a predominately mafic crust below. Such a relation also implies a 100 km thick thermal lithosphere beneath the Altiplano and with a lithospheric thinning of a few tens of kilometers beneath the Puna. Absence of expected increase in lithospheric thickness in regions of almost doubled crust strongly suggests partial removal of the mantle lithosphere beneath the entire plateau. In the Subandean ranges at 19^20‡S, the relation between altitude and crustal thickness indicates a thick lithosphere (up to 130^150 km) and lithospheric flexure. Beneath a relative topographic low at the Salar de Atacama, a thick crust (67 km) suggests that the lithosphere in this region is abnormally cold and dynamically subsided, possibly due to coupling with the subducting plate. This may be related to the strongest (Ms = 8.0) known intra-slab earthquake in the central Andes that happened very close to this region in 1950. The average crustal Vp /Vs ratio is about 1.77 for the Altiplano^Puna and it reaches the highest values (1.80^1.85) beneath the volcanic arc, indicating high ambient crustal temperatures and wide-spread intra-crustal melting. ? 2002 Elsevier Science B.V. All rights reserved. Keywords: Central Andes; Mohorovicic discontinuity; mantle; delamination 1. Introduction The Andes are widely considered as a typical * Corresponding author. Tel.: +49-331-288-1246; Fax: +49-331-288-1277. E-mail address: [email protected] (X. Yuan). orogenic belt, formed due to subduction of an oceanic plate (Nazca plate) under a continental upper plate (South American plate) (e.g. [1]). The central Andean plateau, bounded to the west by the Western Cordillera volcanic arc and to the east by the Eastern Cordillera thrust-fault belt, comprises the Altiplano plateau in the north and the Puna plateau in the south (Fig. 1), and is 0012-821X / 02 / $ ^ see front matter ? 2002 Elsevier Science B.V. All rights reserved. PII: S 0 0 1 2 - 8 2 1 X ( 0 2 ) 0 0 5 8 9 - 7 EPSL 6204 21-5-02 Cyaan Magenta Geel Zwart 390 X. Yuan et al. / Earth and Planetary Science Letters 199 (2002) 389^402 next to the Tibetan plateau in size. The associated combination of subduction and tectonic shortening processes, which resulted in an unusually large crustal thickening at the active margin, is not well understood. Therefore it attracted much attention in the last decade and stimulated a number of geophysical studies to image the deep structure and probe into the processes of the subduction system. Detailed information on the variations of crustal properties and thicknesses is crucial for understanding the mechanism of crustal thickening. Previous geophysical data have shown that the overriding plate has a Moho at 30^80 km depth in some places with its deepest parts beneath the Eastern and Western cordilleras. One of the earliest comprehensive studies of crustal thickness was that of James [2], who estimated the maximum thickness in excess of 70 km beneath the Western Cordillera, based on surface wave dispersion data. The wide-angle re£ection and refraction data [3^6] indicate Moho re£ections beneath the forearc and backarc areas, which normally could not be detected beneath the volcanic arc. This may be due to a transitional Moho and/or due to the highly attenuating nature of crust there. Seismological investigations have detected the continental Moho across the entire plateau, using teleseismic or local earthquake receiver function analysis, regional waveform modeling, and underside Moho re£ection studies [7^10]. Most of these seismic studies have concluded that the average velocity of the crust beneath the plateau is low (Vp = 5.9^6.2 km/s, see [4] and [9]), and its average Poisson’s ratio is about 0.25 (Vp /Vs = 1.73) [8,9,11,12]. This was interpreted as an evidence for the felsic composition of the crust [7^9,11,12]. The teleseismic receiver function technique is now routinely used to investigate the crustal and upper mantle discontinuities beneath permanent and mobile seismic networks. The technique enables generation of high-resolution structural images of the lithosphere and the entire upper mantle, similar to those obtained for the crust by the near vertical re£ection method. A steeply incident P-wave from a distant earthquake penetrates the underlying upper mantle and the crust, before it reaches a seismic station. At each discontinuity the P-wave generates an S-wave that follows it to the station. This converted S-wave is usually very weak, but can be identi¢ed by its delay and polarization if a signi¢cant number of records are available. Using a modi¢cation of the receiver function method, Yuan et al. [13] have shown that the continental Moho is generally seen beneath the central Andes at depths of 40^80 km, with large changes in the north^south direction, thinning by 10^20 km from the Altiplano to the Puna. In this paper we will focus on mapping the Moho topography in the central Andes in greater detail. 2. Data and method Several temporary arrays of seismic stations were operated in northern Chile, southern Bolivia and northwestern Argentina (Fig. 1) for time periods ranging from 2 months to more than 1 yr, supported by the Sonderforschungsbereich (SFB) 267 of the Deutsche Forschungsgemeinschaft and the US PASSCAL project in cooperation with South American institutions (see [13] for details). Most of the stations were equipped with shortperiod 1-Hz Mark-L4 seismometers (black triangles in Fig. 1) while a small number were equipped with Guralp-3T and Streckeisen STS-2 broadband sensors (squares in Fig. 1). The experiments were designed to improve the images of local seismicity, seismic velocity structure and seismic attenuation structure in the central Andes. Besides the local earthquake recordings mostly from the subducted Nazca plate, a number of teleseismic earthquakes and deep regional events were recorded within the operation periods and have been used in our receiver function study. P-waveform data with high signal/noise ratio are selected at each station from the teleseismic earthquake recordings with epicentral distances between 30 and 95‡. Because of the global earthquake distribution pattern and the short observation periods, relatively few useful teleseismic earthquakes could be recorded. Receiver functions for each earthquake^station pair were calculated in the way described by Yuan et al. [14]. The response of the short-period instruments was EPSL 6204 21-5-02 Cyaan Magenta Geel Zwart X. Yuan et al. / Earth and Planetary Science Letters 199 (2002) 389^402 391 Fig. 1. Topographic map of the central Andes with major tectonic features indicated. Small triangles are recent volcanoes. Seismic stations are denoted by the the large triangles (short-period stations, operated for about 3 months) and the black squares (broadband stations, operated for more than 1 yr). The entire area is subdivided into 1U1‡ grids marked by crosses, which are labeled in alphabetical and numerical order shown at the left and bottom. broadened by deconvolving their instrument response. Three-component seismograms were then rotated into a ray-oriented coordinate system in directions of P-, SV- and SH-waves. The rotation angles (back azimuth and incidence angle) were determined by the eigenvalues of the covariance matrix over a time window spanning the ¢rst few seconds following the P-wave arrival. Receiver functions were computed by deconvolving the P-waveforms from the corresponding SV-wave components. In total 642 receiver functions were obtained for more than 170 stations used in this study. It is well known that a relatively large trade-o¡ exists between the velocity^depth estimation. However, since receiver function analysis uses the di¡erential travel times between the P-to-S converted (Ps) waves and the incident P-wave, they are less sensitive to the absolute P- and S-wave velocities compared to the Vp /Vs ratio. Although the Ps-waves at the Moho are usually the strongest phases in receiver functions, multiply re£ected waves between the surface and the Moho (Pps and Pss) are often observed. Zandt et al. [15] have demonstrated that multiple reverberations within the crust, if well observed, can be used together with directly converted waves to accurately constrain the Vp /Vs ratio and the crustal thickness. The Ps-waves at the continental Moho beneath the central Andes and their multiples are well observed in the receiver function data (see also [13] and its supplements), providing a good opportunity to determine the Moho depth and average Vp /Vs ratio for the crust. For this purpose, the grid-search method described by Zhu and Kanamori [16] is adopted, which relies on stacking of receiver functions for varying Moho depth and mean crustal Vp /Vs ratio. The best estimates of the Moho depth and the Vp /Vs ratio are found EPSL 6204 21-5-02 Cyaan Magenta Geel Zwart 392 X. Yuan et al. / Earth and Planetary Science Letters 199 (2002) 389^402 when the Moho conversion and their multiples are stacked coherently. This is illustrated in Fig. 2 for a single broadband station in the area B4 (Fig. 1) located in the volcanic arc of the Western Cordillera in northern Chile, which was operated for more than 1 yr. Note that wide-angle seismic studies [4] could not identify the Moho in this region. In Fig. 2a receiver function data (20 traces) are displayed equally spaced and ordered by their back azimuth. The Ps-phase at the Moho (labeled Ps) and the two major multiple re£ections at the Moho (labelled Pps and Pss, respectively) can be correlated among the individual receiver functions. We summed the individual traces after applying moveout correction for the Ps and Pps phases, respectively (the two traces at the top of Fig. 2a) and found that both the Ps and the Pps phases have been optimally coherently enhanced. Their arrival times are 7.9 and 25.7 s, respectively, after the moveout correction for a reference slowness of 6.4 s/‡. Using the method proposed by Zandt et al. [15] we estimate a Vp /Vs ratio of 1.80 beneath this station. Fig. 2b shows the results of applying the gridsearch algorithm of Zhu and Kanamori [16] to the receiver functions for this station. For each parameter pair of the Moho depth and the crustal average Vp /Vs ratio, travel times of the Ps and the multiples Pps and Pss phases have been calculated. Amplitudes of each receiver function trace corresponding to these calculated travel times of the three phases are stacked. The optimum values of the Moho depth and average Vp /Vs ratio correspond to the maximum energy of stacked phases (58 km Moho depth and 1.80 average crustal Vp /Vs ratio, in agreement with the above estimated values). C Fig. 2. (a) Individual receiver functions of a broadband station located in grid B4. Receiver function data in a window of 310 to 50 s are displayed equally spaced and sorted by back azimuth (BAZ), which is indicated to the left of each trace. Vertical straight lines mark the directly converted phase and two multiple phases, labeled by Ps, Pps and Pss, respectively. The top two traces are summations after moveout correction for the Ps and Pps phases, respectively, have been applied to the individual receiver functions. (b) Gridsearch result for di¡erent Moho depths and crustal average Vp /Vs ratios for receiver functions at the station. High energy (black and dark gray colors) corresponds to the values of the Moho depths and the average Vp /Vs ratios in the crust which optimally ¢t the observed P-to-S conversion at the Moho and its multiples. The white circle indicates the best ¢t. EPSL 6204 21-5-02 Cyaan Magenta Geel Zwart X. Yuan et al. / Earth and Planetary Science Letters 199 (2002) 389^402 Fig. 3. Synthetic receiver functions displayed in the similar way as in Fig. 2. Di¡erence is that the individual receiver functions are sorted by epicentral distances in order to show clearly the moveout for di¡erent phases of Ps, Pps and Pss. The e¡ect of the moveout corrections for di¡erent waves can be seen in the corresponding summation traces. 393 To understand the moveout correction of the Ps multiples and to demonstrate the estimations of the Vp /Vs ratio and the crustal thickness using the grid-search algorithm, we calculated synthetic receiver functions using re£ectivity method [17] for a single layer crustal model with an average Vp of 6.1 km/s, Vp /Vs ratio of 1.80 and crustal thickness of 58 km (Fig. 3a). We presented the data as in Fig. 2, but with a small di¡erence. Unlike those sorted by azimuth, the individual receiver functions in Fig. 3a are sorted by epicentral distance in order to show the di¡erence in moveout behavior of Ps and its multiples. The summation of moveout corrected traces enhances coherently those phases, which have been used to calculate the moveout correction. Other phases are suppressed. The grid-search diagram in Fig. 3b gives exactly the same estimates of the Vp /Vs ratio and the crustal thickness as the real data in Fig. 2. A set of synthetic receiver functions has been calculated for a slowness of 6.4 s/‡ to test the sensitivity of the receiver function data to variations in Vp /Vs ratio and crustal thickness (Fig. 4), using the re£ectivity method [17]. For simplicity, we used a model with a homogeneous crust. The model of Fig. 3 was used for reference. We varied the Vp /Vs ratio between 1.70 and 1.90 and the crustal thickness between 48 and 68 km and calculated synthetic receiver functions for each model (Fig. 4a,b). Increase in both the Vp /Vs ratio and in the crustal thickness delays the travel times of all the phases in the receiver functions. The delay of the Pps phase caused by the increase in the Vp /Vs ratio equals the delay of the Ps phase (Fig. 4a), while the delay of the Pps phase caused by the increase in the crustal thickness is more signi¢cant than that of the Ps phase (Fig. 4b). In Fig. 4c we selected some combinations of Vp /Vs ratios and crustal thicknesses by keeping the time of the Ps phase ¢xed (at 7.9 s). The time variations of the multiple phases are very signi¢cant in response to the parameter variations. If only the Ps phase is observed, the error in estimating the crustal thickness is V14 km for Vp /Vs ratios in the range 1.70^1.90. However, if the multiple phases are also clearly observed, they EPSL 6204 21-5-02 Cyaan Magenta Geel Zwart 394 X. Yuan et al. / Earth and Planetary Science Letters 199 (2002) 389^402 can be very useful to determine the Vp /Vs ratio and the crustal thickness more accurately. In order to apply the grid-search method to the entire data set of the receiver functions, the study region was subdivided into grids of 1U1‡ size. They are labeled in alphabetical and numerical order shown at the left and bottom of Fig. 1. In each grid, we use directly converted and multiply re£ected energy and estimate the average Moho depth and the Vp /Vs ratio. Sums of all individual receiver functions within each grid are plotted in Fig. 5. The traces have been sorted by increasing Fig. 4. Synthetic receiver functions to test the sensitivity to di¡erent model parameters. The reference model has a single-layer crust with crustal thickness of 58 km, P-wave velocity of 6.1km/s, Vp /Vs ratio of 1.80. The synthetics were calculated for a slowness of 6.4 s/‡. (a) Synthetic receiver functions for di¡erent Vp /Vs ratios between 1.70 and 1.90. (b) Synthetic receiver functions for di¡erent crustal thicknesses between 48 km and 68 km. (c) Synthetic receiver functions for di¡erent combinations of the Vp /Vs ratios and the crustal thicknesses. The arrival time of the Ps-wave in each receiver function was kept unchanged. Fig. 5. Summations of receiver functions in each 1U1‡ grid, displayed in the order of their Ps conversion times. Di¡erent ¢lters were used to display the Moho Ps and the multiples. While the Ps window (between 0 and 15 s) is broadband, the multiples window (between 15 and 40 s) has been 5 s lowpass ¢ltered. Moveout corrections have been applied correspondingly in each window prior to summation. Only data showing clear Pps multiples were used for the Vp /Vs estimates. Other multiples are not displayed. For some traces the Pss multiples can be also seen. The corresponding grid index is marked at the left of each trace. The numbers to the right of each trace denote the number of receiver functions stacked. Phase arrivals are marked with white squares. EPSL 6204 21-5-02 Cyaan Magenta Geel Zwart X. Yuan et al. / Earth and Planetary Science Letters 199 (2002) 389^402 times of Ps conversions of the continental Moho. Because the multiples travel three times longer in the crust compared to the Ps-waves, they are usually weak and scattered. The data in Fig. 5 are split into two windows. The ¢rst 15 s window contains mainly the energy from the Moho conversions. The second window between 15 and 40 s contains Moho multiples and has been low-pass ¢ltered with a 5 s cut-o¡ period. Moveout corrections have been applied corresponding to the individual receiver functions in each window prior to summation. Only data showing clear Pps multiples have been used for the Vp /Vs estimates which are displayed in the second window. In some traces the Pss multiples can be also seen. The large amplitude peaks on most traces directly after the P arrival at 0 s are probably caused by interference between converted energy at shallow depth and some P energy remaining from imperfect rotation of the coordinate system. 3. Results The grid-search algorithm has been applied for receiver functions at stations within each grid in Fig. 1. The method works generally well for most grids, but at some grids the Moho conversions are di⁄cult to identify. For A1, B1, B2 and C1, strong converted energy of the oceanic Moho dominates the receiver functions and masks the continental Moho conversion. The data at D3 are quite noisy, preventing us from obtaining a stable Moho stack. The two easternmost stations in the Chaco plain have been summed together to estimate the Moho depth (F9). The resulting map of Moho topography is shown in Fig. 6a. The Moho depth varies from 35 km in the forearc to more than 70 km beneath the plateau and it becomes thinner (30 km) further to the east in the Chaco plain. Our results are very close to the results of previous receiver function studies [7,10], and also to the results of wideangle re£ection studies [4,6]. From our additional new data we also see a clear variation of the Moho depth between the northern and southern parts of the region. North of 23‡S the Moho is generally deeper than 60 km with its deepest part 395 beneath the Eastern and Western cordilleras (more than 70 km at E5, C5, E3 and G5), while south of 23‡S it is generally shallower than 60 km (in A4 it is as shallow as 42 km), although the average altitude in the south is higher than in the north. Note that the tendency of the Moho depth to decrease southwards was previously reported from the few wide-angle data and is now fully con¢rmed with our observations using a di¡erent method. Another interesting result is the unexpected deep Moho (67 km) beneath the Salar de Atacama (B3), which has a much lower elevation than the adjacent mountain regions. The Vp /Vs ratios obtained from the grid-searching procedure are shown in Fig. 6b. Compared to the Moho depth estimates, the Vp /Vs ratio estimates are less robust. Nevertheless a general pattern in variation of the ratio can be seen. Most of the Vp /Vs values are rather high in accordance with Yuan et al. [13]. The highest Vp /Vs ratios are mostly observed beneath the volcanic arc (Fig. 6b). Note that neither the Moho depths nor Vp /Vs ratios could be estimated at the Altiplano^Puna volcanic complex (APVC, marked by a red curve in Fig. 6b) and at the volcanic arc, immediately to the north of the APVC. This is due to the extremely energetic conversions from the low-velocity zone in the middle crust and their multiples [12,15] which tend to mask the Moho conversions. At this point, we discuss the errors associated with the Moho depth and Vp /Vs determinations (shown in Fig. 6) obtained by the method of Zhu and Kanamori [16]. We compare these results with travel time observations of the Ps and Pps phases in Fig. 5. From these phases alone (see Fig. 3) we can uniquely determine the average Moho depth and Vp /Vs ratio for a homogeneous crust. The errors depend on the reading accuracy, provided the phases are correctly identi¢ed and the signal^noise ratio is good enough. It can be seen in Fig. 5 that the reading error of Ps (marked by squares) is a small fraction of a second (0.1^0.2 s), whereas the errors of Pps (upright triangles) are about 0.5 s, although we have selected only the traces with the best available multiples. Sorting the traces (in Fig. 5) with increasing Moho conversion time makes it also easier to identify EPSL 6204 21-5-02 Cyaan Magenta Geel Zwart 396 X. Yuan et al. / Earth and Planetary Science Letters 199 (2002) 389^402 the multiples by considering the whole suite of seismograms. We disregard the error of Ps in comparison with the larger errors of Pps and estimate from Fig. 4c the errors of the Moho depth and Vp /Vs due to the estimated U 0.5 s error of Pps. It follows that the errors of the data in Fig. 6 resulting from the inaccurate readings of the multiples are about 0.02 for Vp /Vs and U 1.5 km for the Moho depth, with Moho conversion times kept ¢xed. These errors are about three times EPSL 6204 21-5-02 Cyaan Magenta Geel Zwart X. Yuan et al. / Earth and Planetary Science Letters 199 (2002) 389^402 397 Fig. 6. (a) Map of the Moho depth estimated from the P-to-S conversion at the Moho and its multiples by the grid-search method. The Moho depths are color-scaled and also indicated in each grid with numbers (in km). Shaded grids indicate where the Moho depth is estimated by primary and multiple phases and therefore an average crustal Vp /Vs ratio is available. For the unshaded grids the Vp /Vs ratio is set to 1.73 for estimating the Moho depth by the primary converted phase alone. White grids with numbers show Moho depth estimated from near-critical re£ections. We note good consistency between the receiver function and wide-angle re£ection results. (b) Map of the average Vp /Vs ratio in the crust estimated from the P-to-S conversion at the Moho and its multiples by the grid-search method. The highest Vp /Vs values are observed close to the volcanic arc. Data at grids D3, D4 and especially C3 and C4 (indicated by a red contour) are strongly dominated by the negative conversions at middle crustal depths (see [13] and [18]) which do not allow a reliable estimate of either the Moho depth or the average Vp /Vs ratio but most likely indicate partially molten crust. 6 smaller than the increments in the color scales chosen in Fig. 6. In the following we discuss some extreme values in Fig. 6. The Moho depth at the neighboring grids E5 (80 km) and F4 (57 km) is quite di¡erent. The Ps conversion in F4 is good (Fig. 5). It is not that clear in E5, because there are several more oscillations following the identi¢ed Moho phase, and E5 has only four traces. The Pps multiple in E5 looks well, whereas Pps in F4 has relatively large noise preceeding it. The other Moho conversions in the vicinity of E5 (F3 and G5) are also good, con¢rming a region of deep Moho there, with large depth variation over short distances. F4 is the only grid in the Altiplano region with a Moho depth less than 60 km, the largest Moho depths being 70^80 km. Swenson et al. [11,12] reported Moho depths of 60^65 km under the Altiplano using regional earthquake records. The lateral resolution of the receiver function method is certainly better than that of any wide-angle method. The Moho depths at grids A2^A6 seems to oscillate. The suite of traces sampling A2, A4, F8 and A6 (Fig. 5) varies homogeneously, leading to a steady increase of Moho depth from A2 over A4 to A6. The Moho conversion times of A3 and A5 are later. Most Moho signals are clear in the A sequence, except that A2 and A3 have double signals, making these signals less certain. However, we believe that the change in Moho depth at 24‡S by up to 15 km over short distances is a robust result. Another clear result is that all the grids in the Puna have a Moho depth less than 60 km. Vp /Vs of F8 is questionable since it does not ¢t in the homogeneous sequence of Ps conversions and Pps multiples in A2, A4, F8 and A6. It should rather be between 1.75 and 1.80 according to the travel time observations alone. Such a value would also agree better with neighboring grids F7 and F9. The reason for this problem is probably that the optimum amplitude summation method of Zhu and Kanamori [16] considers a very late Pss of F8, due to noise. A Vp /Vs ratio of 1.73 has been reported by Swenson et al. [11,12] under the entire Altiplano. This more or less agrees with most of our results from the BANJO and SEDA stations, but clearly disagrees with our results at grids F3 and F4. The large Vp / Vs in F3 and F4 depends on the early Pps arrival (Fig. 5). In order to be consistent with a Vp /Vs of 1.73, the Pps phase must arrive 2 and 3 s later in F3 and F4, respectively, than indicated in Fig. 5. This seems to be unlikely. 4. Discussion Here we analyze the estimated crustal thickness in conjunction with the surface topography and gravity and discuss the possible geodynamic consequences. Fig. 7a shows the average topography vs. Moho depth in each 1U1‡ grid. Blue diamonds correspond to the grids north of the 22‡S and red diamonds to the grids south of the 22‡S. The most interesting feature in the Fig. 7a is a poor correlation between surface topography and Moho depth below the Altiplano and Puna. Although the Moho depth varies by more than 20 km, there is no considerable variation of the surface topography. Fig. 7b shows that the same is true for the Bouguer gravity; variable crustal thickness does not result in signi¢cant variations in Bouguer gravity under the high plateaus. This means that either (i) mass de¢cits (excesses) due to EPSL 6204 21-5-02 Cyaan Magenta Geel Zwart 398 X. Yuan et al. / Earth and Planetary Science Letters 199 (2002) 389^402 Fig. 7. (a) Average altitudes versus depths to the Moho in the 1U1‡ grids. Blue diamonds correspond to the grids north of 22‡S and red diamonds to the grids south of 22‡S. Dotted line shows calculated isostatic altitude^Moho depth relations for the crust composed from felsic rocks. Solid line, labelled Hl = 100 km, shows the best ¢t of the data using a two layer lower crust composed of felsic and ma¢c rocks. Curve shape is mostly controlled by the crustal composition. The curve is shifted parallel to the vertical (altitude) axis with changes in thermal lithospheric thickness. Solid curves show expected altitude^Moho depth relations for the mixed felsic^ma¢c lower crust for 50, 100 and 150 km lithospheric thickness. Most of the data are consistent with mixed felsic^ma¢c composition of the lower crust and lithospheric thickness about 100 km. However, there are four grids with large deviations from the average curve. Corresponding points are outlined by the circles. (b) Bouguer gravity anomalies [33] versus depths to the Moho in 1U1‡ grids. variable Moho depth are compensated by excesses (de¢cits) of lithospheric density or (ii) the lowermost crust beneath the plateaus has density close to the mantle and thus changes in thickness do not produce signi¢cant gravity or topographic anomalies. Let us consider ¢rst case (i) and assume that the entire crust beneath the plateaus is felsic [8]. We use the thermodynamic modelling approach of Sobolev and Babeyko [19] to calculate mineral equilibrium, densities and seismic velocities for the central Andean rocks with bulk chemical compositions taken from Lucassen et al. [20]. The thin dotted line in Fig. 7 shows the calculated isostatic altitude^Moho depth relation for the crust composed from average central Andean felsic rocks after [20] and an Altiplano geotherm suggested in [21] (the results are the same for another Altiplano geotherm [22]). This curve is consistent with our data for crustal thickness smaller than 50^55 km, but completely inconsistent with a thicker crust. Topography of the lithosphere^asthenosphere boundary could compensate this mis¢t. However, to do so the lithospheric thickness must vary by more than 200 km, which is unlikely. The observations, however, can be readily explained if the lowermost crust below the plateau has predominantly ma¢c composition (some 80 vol.% of ma¢c rock and 20 vol.% of felsic rock) with average density about 3.2 g/cm3 (solid curve in Fig. 7a). Average P-wave velocity of such lower crustal layer is estimated to be about 7.3 km/s at temperatures above 1000‡C, expected at a depth larger than 50^55 km below Altiplano [21,22]. For comparison, pure felsic lowermost crust at these conditions would have a P-wave velocity of about 6.6 km/s, owing to beta quartz stability. Average velocities of the pure felsic crust and the felsic crust with a 10 km thick layer of predominantly ma¢c rocks at the crustal base would di¡er by only 0.09 km/s, a di¡erence which is hardly detectable by relatively low-resolution seismological methods. Summarizing this part of the discussion we emphasize that our data suggest that the crust below Altiplano and Puna is felsic to the depth of 50^55 km but has more ma¢c composition below this depth. This may suggest that the thick crust in the high-plateau region has achieved its critical thickness. Further thickening must have generated high-grade metamorphic rocks with a EPSL 6204 21-5-02 Cyaan Magenta Geel Zwart X. Yuan et al. / Earth and Planetary Science Letters 199 (2002) 389^402 density higher than the upper mantle, leading to their delamination and detachment [23,24]. If correct, this means that a part of the high-plateau crust could have been lost during shortening. It is interesting that ma¢c lowermost crust is present mostly beneath the Altiplano and is almost absent below the Puna (with the exception of the grid C5) in accordance with mantle delamination hypotheses suggested for the Puna [24]. An interesting question is whether a signi¢cant amount of partial melt is present in the Altiplano^ Puna crust or not. Swenson et al. [12] have concluded that there is no evidence for the partial melting in the Altiplano crust after comparing their estimated average Poisson’s ratio (0.25) and Vp estimates (6.0 km/s) in the crust, with those expected for a felsic crust. However, possible beta quartz stability in the hot crust of the Altiplano is not accounted for in their calculation. We have calculated the expected average Vp and Vp /Vs ratios (Poisson’s ratio) for a 60 km thick melt-free crust of felsic composition for convective and conductive temperature distributions in the crust of Altiplano. The ‘conductive’ model [22] does not take into account possible convective heat transfer by the partially molten or close to solidus felsic rocks and therefore generates very high temperatures in the crust in an attempt to ¢t high surface heat £ow observed in Altiplano. The ‘convective’ model [21] is based on thermo-mechanical modelling of the tectonically shortened felsic crust and it predicts temperatures in the lower crust lower than the ‘conductive’ model. In our calculation we take into account alpha^ beta quartz transition and associated non-linear changes of elastic moduli close to the transition [19]. The estimated average Vp appears to be 6.22 km/s for the ‘conductive’ and 6.17 km/s for the ‘convective’ temperature models. The corresponding Vp /Vs ratios are close to 1.68 (Poisson’s ratio 0.23). From these calculations it is clear that in order to achieve an average Vp of 6.00 km/s [12] a small degree of partial melt is required even for the purely felsic crust. The same conclusion holds for the crust with a thin ma¢c layer at its base, because the resulting average velocity and Vp /Vs ratio is almost the same (6.31 Rm/s, 1.68 for the conductive and 6.26 Rm/s, 1.68 for the convective 399 model). Depending on melt geometry, presence of a small degree of partial melt decreases Vp and increases Vp /Vs ratio. Using the model by Watanabe [25], we obtain the amount of partial melt required to decrease Vp by 4^5% (from 6.25^6.3 to 6.0 km/s) as 3^4 vol.% with a simultaneous increase of Vp /Vs ratio by some 4%, reaching 1.75 (0.26). This is just in-between our estimate of the average Vp /Vs ratio for the crust in Altiplano and Puna (1.77) and that in [11,12] (1.73). Note, that high Vp /Vs ratios obtained by us correlate well with the pronounced low-velocity zone in the middle crust (20^40 km depth), detected by both receiver functions [13] and wide-angle re£ections [4] in Altiplano, suggesting that partial melt is mostly concentrated in the middle crust, rather than in the lower crust. This result is in good agreement with the results of thermo-mechanical modelling of the Altiplano crust [21]. Thickness of the lithosphere (crustal plus mantle parts) is another factor, which a¡ects the isostatic equilibrium altitude of the crust. If only the lithospheric thickness is varied keeping crustal densities ¢xed, then the shape of the altitude^ Moho depth curve (solid line in Fig. 7a) does not change, but this curve shifts parallel to the vertical (altitude) axis. To a ¢rst approximation, we can quantify this e¡ect using data from other, better studied regions of the world. Lithospheric thickness in the volcanic regions of the French Massif Central is estimated to be about 50^60 km [26] from interpretation of the high-resolution tomographic models. Average altitude (slightly above 1 km) and crustal thickness (close to 30 km) [27] ¢t the curve labeled Hl = 50 km in Fig. 7a. Lithospheric thickness is around 100 km away from the volcanic ¢elds of the French Massif Central [26], which might be typical for tectonically active western Europe [27]. Taking 28^30 km as typical crustal thickness and 0^0.2 km as the altitude for active western Europe we ¢x the curve corresponding to a lithospheric thickness of 100 km, which is labeled Hl = 100 km in Fig. 7a. Continuing this procedure linearly we obtain the curve labeled Hl = 150 km which corresponds to the 150 km thick lithosphere. These estimates are rather crude, but nevertheless suggest that the data for the high-plateau are con- EPSL 6204 21-5-02 Cyaan Magenta Geel Zwart 400 X. Yuan et al. / Earth and Planetary Science Letters 199 (2002) 389^402 sistent with a lithospheric thickness less than 100 km rather than 150 km. This conclusion is important because due to strong tectonic shortening, the lithosphere must almost be doubled in the high-plateau region. This means that either the lithosphere had to be very thin (some 50 km) before shortening or a signi¢cant part of the lithospheric mantle had to be removed during shortening. As there is no evidence for a very thin lithosphere in the entire high-plateau region before shortening, we consider lithospheric mantle removal to be the most plausible explanation. Fig. 8 shows deviations of the observed altitudes from the theoretical curve for the crust with a ma¢c layer. With a few exceptions, which will be discussed below, the deviations are small, signi¢cantly less than 0.5 km. There is a small systematic di¡erence between the Puna (south of 22‡S) and Altiplano (north of 22‡S), also clearly visible in Fig. 7a,b, which can be explained either by some 20 km thinner lithosphere beneath the Puna or by a less dense Puna crust. Analysis of Cenozoic magmatism [24], as well as evidence from lower seismic velocities and higher attenuation in the uppermost Fig. 8. Di¡erence between observed altitude and theoretically expected altitude. The largest negative deviations (blue ^ theoretical altitude is too high) are located in the Subandean ranges (F6, F7) and in the Salar de Atacama block (B3). The largest positive is in the Puna (A4). Also shown are the grid-search results for four grids with largest di¡erences between observed and expected altitude. EPSL 6204 21-5-02 Cyaan Magenta Geel Zwart X. Yuan et al. / Earth and Planetary Science Letters 199 (2002) 389^402 mantle beneath the Puna than beneath the Altiplano [28], support the idea of a relatively thinner lithosphere beneath the Puna. Although most of the data ¢t well with the theoretical curve shown in Fig. 7a, there are several points which do not. They are outlined by the circles in Fig. 7. The corresponding 1U1‡ grids show up as strong blue or red colors in Fig. 8. As also demonstrated in Fig. 8, the grid-search method provides good estimations for the crustal thickness for all these anomalous crustal blocks. Therefore, these deviations likely re£ect anomalous deep structure and dynamic state rather than artifacts of our analysis technique. The elevation anomaly in the Subandes (grids F7^F8, east of 63‡W) can only be partially caused by £exure of the Brazilian lithosphere, which is underthrusting the Andean crust [29]. The rest of the anomaly could be explained by some 30^50 km thicker lithosphere in the Subandes than beneath the high-plateau. This is consistent with the higher seismic velocity and lower attenuation in the Subandean upper mantle than in the adjacent Eastern Cordillera at 20‡S [30]. Of particular interest is the thick (67 km) crust observed in the Salar de Atacama block (B3), whose altitude is about 2.5 km. This is more than 1 km below the value expected from a crustal thickness of 67 km if isostatic equilibrium is achieved, and it has anomalously high Bouguer gravity. Elevation of this block would be consistent with the 150 km thick lithosphere (see Fig. 7a), but this could not be the case because the subducting Nazca plate is located at only 100 km depth below this block. From this discrepancy and also from seismic tomographic data showing high seismic velocities and very low attenuation in the mantle below this block [28], we infer that the Salar de Atacama block has an abnormally cold lithospheric mantle which is mechanically coupled with the top of the subducting Nazca plate. The coupling between this cold block (acting as a giant asperity) and the slab might be the reason that the strongest (Ms = 8) known intra-slab earthquake of 1950 in the central Andes [31] occurred close to the Salar de Atacama block. Another anomalous block (A4) is located close to the Salar de Atacama block to the other side of 401 the volcanic arc, and it stands higher than would be expected from its crustal thickness of 42 km if isostatic equilibrium is achieved. As seen from Fig. 7a, the observed elevation of this block (about 4 km) can not be attained even if the ambient crust is purely felsic and the lithosphere is very thin (50 km). Therefore, we expect that this block may be in non-isostatic equilibrium. It is interesting that the average elevation and Bouguer gravity of the two adjacent anomalous blocks, B3 and A4, treated as one joint block, is very close to that expected for their average crustal thickness. If those two blocks were mechanically coupled, the equilibrium for such a combined heterogeneous block would require block B3 to be dynamically subsided and block A4 to be uplifted [32], in accordance with our observations. Acknowledgements We thank R. Kumar, B. Schurr and D. Hindle for carefully reading the manuscript and H.J. Go«tze for providing Bouguer gravity data. The ¢eld experiments have been supported by the SFB 267 of the Deutsche Forschungsgemeinschaft, the GeoForschungsZentrum Potsdam, the Freie Universita«t Berlin, the US National Science Foundation, the BANJO and SEDA projects, the Universidad de Chile (Santiago), the Universidad Catolica del Norte (Antofagasta) and the Universidad National Salta.[RV] References [1] R.W. Allmendinger, T.E. Jordan, S.M. Kay, B.L. Isacks, The evolution of the Altiplano^Puna Plateau of the Central Andes, Annu. Rev. Earth Planet. Sci. 26 (1997) 139^ 174. [2] D.E. James, Andean crustal and upper mantle structure, J. Geophys. Res. 76 (1971) 3246^3271. [3] P.J. Wigger, Seismicity and crustal structure of the Central Andes, in: H. Bahlburg, Ch. Breitkreuz, P. Giese (Eds.), The Southern Central Andes, Springer Verlag, Berlin, 1988, pp. 209^229. [4] P.J. Wigger, M. Schmitz, M. Araneda, G. Asch, S. Baldzuhn, P. Giese, W.-D. Heinsohn, E. Mart|¤nez, E. Ricaldi, P. Ro«wer, J. Viramonte, Variation in the crustal structure of the southern central Andes deduced from seismic re- EPSL 6204 21-5-02 Cyaan Magenta Geel Zwart 402 [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] X. Yuan et al. / Earth and Planetary Science Letters 199 (2002) 389^402 fraction investigations, in: K.-J. Reutter, E. Scheuber, P.J. Wigger (Eds.), Tectonics of the Southern Central Andes, Springer Verlag, Berlin, 1994, pp. 23^48. R. Patzwahl, J. Mechie, A. Schultze, P. Giese, Two-dimensional velocity models of the Nazca plate subduction zone between 19.5‡S and 25‡ from wide-angle seismic measurements during the CINCA95 project, J. Geophys. Res. 104 (1999) 7293^7317. M. Schmitz, K. Lessel, P. Giese, P. Wigger, M. Araneda, J. Bribach, F. Graeber, S. Grunewald, C. Haberland, S. Lueth, P. Roewer, T. Ryberg, A. Schulze, The crustal structure beneath the Central Andean forearc and magmatic arc as derived from seismic studies ^ the PISCO 94 experiment in northern Chile (21‡^23‡S), J. South Am. Earth Sci. 12 (1999) 237^260. S.L. Beck, G. Zandt, S.C. Myers, T.C. Wallace, P.G. Silver, L. Drake, Crustal thickness variations in the Central Andes, Geology 24 (1996) 407^410. G. Zandt, A.A. Velasco, S.L. Beck, Composition and thickness of the southern Altiplano crust, Bolivia, Geology 22 (1994) 1003^1006. G. Zandt, S.L. Beck, S.R. Ruppert, C.J. Ammon, D. Rock, E. Minaya, T.C. Wallace, P.G. Silver, Anomalous crust of the Bolivian Altiplano, Central Andes: constraints from broadband regional seismic waveforms, Geophys. Res. Lett. 23 (1996) 1159^1162. G. Bock, B. Schurr, G. Asch, High-resolution image of the oceanic Moho in the subducting Nazca plate from P^S converted waves, Geophys. Res. Lett. 27 (2000) 3929^3932. J.L. Swenson, S.L. Beck, G. Zandt, Regional distance shear-coupled PL propagation within the northern Altiplano, central Andes, Geophys. J. Int. 139 (1999) 743^ 753. J.L. Swenson, S.L. Beck, G. Zandt, Crustal structure of the Altiplano from broadband regional waveform modeling: Implications for the composition of thick continent crust, J. Geophys. Res. 105 (2000) 607^621. X. Yuan, S. Sobolev, R. Kind, O. Oncken, G. Bock, G. Asch, B. Schurr, F. Graeber, A. Rudlo¡, W. Hanka, K. Wylegalla, R. Tibi, C. Haberland, A. Rietbrock, P. Giese, P. Wigger, P. Ro«wer, G. Zandt, S. Beck, T. Wallace, M. Pardo, D. Comte, New constraints on subduction and collision processes in the Central Andes from P-to-S converted seismic phases, Nature 408 (2000) 958^961. X. Yuan, J. Ni, R. Kind, J. Mechie, E. Sandvol, Lithospheric and upper mantle structure of southern Tibet from a seismological passive source experiment, J. Geophys. Res. 102 (1997) 27491^27500. G. Zandt, S.C. Myers, T.C. Wallace, Crustal and mantle structure across the Basin and Range ^ Colorado Plateau boundary at 37‡N latitude and implications for Cenozoic extensional mechanism,, J. Geophys. Res. 100 (1995) 10529^10548. L. Zhu, H. Kanamori, Moho depth variation in southern California from teleseismic receiver functions, J. Geophys. Res. 105 (2000) 2969^2980. [17] R. Kind, The re£ectivity method for a buried source, J. Geophys. 44 (1978) 603^612. [18] J. Chmielowski, G. Zandt, C. Haberland, The central Andean Altiplano^Puna Magma body, Geophys. Res. Lett. 26 (1999) 783^786. [19] S.V. Sobolev, A.Yu. Babeyko, Modelling of mineralogical composition, density and elastic wave velocities in the unhydrous rocks, Surv. Geophys. 15 (1994) 515^544. [20] F. Lucassen, S. Lewerenz, G. Franz, Metamorphism, isotopic ages and composition of lower crustal granulite xenoliths from the Cretaceous Salta Rift, Argentina, Contrib. Mineral. Petrol. 134 (1999) 325^341. [21] A. Yu. Babeyko, S.V. Sobolev, R.B. Trumbull, O. Oncken, L.L. Lavier, Numerical models of crustal scale convection and partial melting beneath the Altiplano^Puna plateau, Earth Planet. Sci. Lett., submitted. [22] M. Springer, Tectonophysics 306 (1999) 377^395. [23] S.V. Sobolev, A.Y. Babeyko, Phase transformations in the lower continental crust and its seismic structure, in: R.S. Mereu, S. Mueller, D.M. Fountain (Eds.), Properties and Processes of Earth’s Lower Crust, Geophys. Monogr. 51 (1989) 311^320. [24] R.W. Kay, S.M. Kay, Delamination and delamination magmatism, Tectonophysics 219 (1993) 177^189. [25] T. Watanabe, E¡ects of water and melt on seismic velocities and their application to characterization of seismic re£ectors, Geophys. Res. Lett. 20 (1993) 2933^ 2936. [26] S.V. Sobolev, H. Zeyen, M. Granet, U. Achauer, C. Bauer, F. Werling, R. Altherr, K. Fuchs, Upper mantle temperatures and lithosphere^asthenosphere system beneath the French Massif Central constrained by seismic gravity, petrologic and thermal observations, Tectonophysics 275 (1997) 143^164. [27] H. Zeyen, F. Volker, V. Wehrly, K. Fuchs, S.V. Sobolev, R. Altherr, Styles of continental rifting crust-mantle detachment and mantle plumes, Tectonophysics 278 (1997) 329^352. [28] B. Schurr, Seismic Structure of the Central Andean Subduction Zone From Local Earthquake Data, Ph.D Thesis, FU, Berlin, 2000. [29] H. Lyon-Caen, P. Molnar, G. Sua¤rez, Gravity anomalies and £exure of the Brazilian shield beneath the Bolivian Andes, Earth Planet. Sci. Let. 75 (1985) 81^92. [30] S. Myers, S. Beck, G. Zandt, T. Wallace, Lithosphericscale structure across the Bolivian Andes from tomographic images of velocity and attenuation for P and S waves, J. Geophys. Res. 103 (1998) 21233^21252. [31] E. Kausel, J. Campos, The Ms = 8 tensional earthquake of 9 December 1950 of northern Chile and its relation, Phys. Earth Planet. Inter. 72 (1992) 220^235. [32] E.V. Artyushkov, Stresses in the lithosphere caused by crustal thickness inhomogeneities, J. Geophys. Res. 78 (1973) 7675^7708. [33] H.-J. Go«tze, A. Kirchner, Gravity ¢eld at the South American active margin (20‡ to 29‡ S), J. South Am. Earth Sci. (1997) 179^188. EPSL 6204 21-5-02 Cyaan Magenta Geel Zwart
© Copyright 2026 Paperzz