Originally published as:

Transcription

1 Originally published as: Heuer, B., Geissler, W. H., Kind, R., BOHEMA working group (2011): Receiver function search for a baby plume in the mantle transition zone beneath the Bohemian Massif. - Geophysical Journal International, 187, pp DOI:

2 Geophysical Journal International Geophys. J. Int. (2011) 187, doi: /j X x Receiver function search for a baby plume in the mantle transition zone beneath the Bohemian Massif Barbara Heuer, 1 Wolfram H. Geissler, 2 Rainer Kind 1 and the BOHEMA working group 1 Deutsches GeoForschungsZentrum Potsdam, Telegrafenberg, Potsdam, Germany. barbara.hahne@liag-hannover.de 2 Alfred Wegener Institute for Polar and Marine Research, Am Alten Hafen 26, Bremerhaven, Germany Accepted 2011 July 13. Received 2011 July 7; in original form 2010 September 22 SUMMARY The western Bohemian Massif is known for geodynamic phenomena such as earthquake swarms, CO 2 dominated free gas emanations of upper-mantle origin, and Tertiary/ Quaternary volcanism. Among other explanations, a small-scale mantle plume has been suggested. We used data from the international passive seismic experiment BOHEMA ( ) and of a previous seismic experiment to investigate the structure of the upper-mantle discontinuities at 410 km and 660 km depth (the 410 and the 660 ) beneath the Bohemian Massif with the P receiver function method. More than 4500 high-quality receiver function traces could be utilized. Two stacking techniques were used: stacking by station (common station method, CSM) and stacking by piercing points in the mantle transition zone (common conversion point method, CCM). Since the station spacing is very close, rays from different stations have similar piercing points in the mantle transition zone. Therefore CCM is sensitive in the transition zone and CSM is sensitive to the uppermost structure of the mantle. The CSM shows delayed conversion times from the 410 km discontinuity beneath the western Bohemia earthquake region, which indicate a slow uppermost mantle. When stacking our data by CCM, we observe thickening of the transition zone towards the Alpine foreland, which agrees with tomographic results by Piromallo and Morelli. The thickness of the mantle transition zone beneath the western Bohemian Massif is normal, with a faint hint to thinning in the northern part. Our conclusion is that a plume-like structure may exist in the upper mantle below the western Bohemia earthquake region, but with no or only weak imprint on the 410 km discontinuity. Key words: Mantle processes; Body waves; Intra-plate processes; Dynamics of lithosphere and mantle. GJI Geodynamics and tectonics Now at: Leibniz Institute for Applied Geophysics, Stilleweg 2, Hannover, Germany. Members of the BOHEMA working group: V. Babuška, J. Plomerová, L. Vecsey, J. Zedník, P. Jedlička, V. Vavryčuk, J. Horálek, A. Boušková, T. Fischer and B. Ržek (Institute of Geophysics, CAS, Prague); M. Brožand J. Málek (Institute of Rock Structure and Mechanics, CAS, Prague); V. Nehybka (Institute of Phys. Earth, Masaryk University, Brno); O. Novotný (Department of Geophysics, Charles University, Prague); M. Granet, U. Achauer and T. Piquet (Institut de Physique du Globe, Univ. Strasbourg); R. Kind, H. Kämpf, W. Geissler and B. Heuer (Deutsches, GeoForschungsZentrum Potsdam); M. Korn, S. Wendt and S. Funke (Institut für Geophysik, Univ. Leipzig); K. Klinge, T. Plenefisch, K. Stammler and M. Lindemann (BGR, Hannover); K. Bräuer (Umweltforschungszentrum Leipzig- Halle); G. Jentzsch, P. Malischewski and M. Brunner (Institut für Geowissenschaften, Univ. Jena). 1 INTRODUCTION The western Bohemian Massif in central Europe is situated at the junction of three Variscan structural units: the Saxothuringian in the north, the Teplá-Barrandian and Moldanubian units in the south (Fig. 1). The Palaeozoic suture between the Saxothuringian and Teplá-Barrandian/Moldanubian units has been reactivated since the Upper Cretaceous/Tertiary as part of the European Cenozoic Rift System. This led to the evolution of the 300 km long and 50 km wide ENE WSW trending Eger (Ohře) Rift. The western part of the Eger Rift is known for geophysical and geological phenomena such as the occurrence of earthquake swarms, CO 2 dominated free gas emanations of subcontinental lithospheric mantle signature in mineral springs and mofettes, Tertiary/Quaternary volcanism and neotectonic crustal movements (e.g. Špičák & Horálek 2001; Ulrych et al. 2003; Bräuer et al. 2005). C 2011 The Authors 577

3 578 B. Heuer et al. Figure 1. Map of the Bohemian Massif with its major tectonometamorphic units and fault zones (modified after Plomerová et al. 2007): Mid-German Crystalline High (MGCH); Saxothuringian (ST), Moldanubian (MD) and Teplá-Barrandian (TB) unit; Mariánské Lázně Complex (MLC); Central Bohemian Pluton (CBP); Eger Graben (EG); West Bohemian Shear Zone (WBSZ); Central Bohemian Shear Zone (CBSZ); Bavarian Shear Zone (BSZ); Elbe Fault Zone (EFZ); Blanice Graben (BLG); Intra-Sudetic Fault (ISF); Regensburg-Leipzig-Rostock zone (RLR). KTB is the location of the German Continental Deep Drilling Boreholes. Black triangles show the location of stations of the BOHEMA experiment and of the prestudy by Geissler et al. (2005). Inset: Location of the investigated area within central Europe. AM: Armorican Massif; MC: Massif Central; RM: Rhenish Massif; BM: Bohemian Massif; ST: Saxothuringian unit; MD: Moldanubian unit; RH: Rhenohercynian unit. To explain the observed phenomena, several possible scenarios have been suggested: a small-scale mantle plume (Granet et al. 1995), lithospheric thinning beneath the Eger Rift (Babuška & Plomerová 2001; Heuer et al. 2006), and presently ongoing magmatic processes near the crust mantle boundary, including magmatic underplating (Bräuer et al. 2005; Geissler et al. 2005). Seismic tomography investigations of the French Massif Central and the Eifel volcanic region (Rhenish Massif) indicated plumelike structures in the upper mantle (Granet et al. 1995; Ritter et al. 2001). These regions are geologically comparable to the western Bohemian Massif as all of them are of Variscan origin and were overprinted by the European Cenozoic Rift system with its Tertiary and Quaternary volcanism and recent CO 2 emanations. However, due to the geometry of the tomography experiments, the plume-like structures could only be resolved down to depths of 270 and 400 km for the French Massif Central and the Eifel region, respectively. Granet et al. (1995) suggested that the hot mantle finger might have its origin in the upper mantle near a thermal boundary layer, like the top of the mantle transition zone at 410 km depth, and predicted other diapiric mantle upwellings beneath the Rhenish Massif, the Bohemian Massif and the Pannonian Basin. Goes et al. Figure 2. Distribution of teleseismic events used in this study. The black star in the centre marks the location of the investigation area. Events that occurred within the recording time span of the BOHEMA experiment are shown by red dots. Events that occurred before and after that time span are coloured black and were recorded by permanent stations and/or by stations of the experiment of Geissler et al. (2005). Generally, there is a gap of observations to the south and dominance of observations to the northeast of the study area. (1999) observed low seismic velocities in the lower mantle beneath central and western Europe and suggested that they may represent a mantle plume and a common source for Cenozoic volcanism in central Europe. However, the relationship of this deep-seated lowvelocity structure to diapiric upwellings in the upper mantle is not yet resolved. Grunewald et al. (2001) and Budweg et al. (2006) carried out receiver function investigations in the Rhenish Massif of western Germany (Eifel region). They further support the idea of a plume beneath the Eifel by their findings of an approximately 200 km wide and 15 to 25 km deep apparent depression of the 410 km discontinuity in the upper mantle beneath the Eifel region. In the western Bohemian Massif, a similar diapiric mantle upwelling was suggested by Granet et al. (1995) and might be plausible due to the similar geological and geodynamic situation as in the French Massif Central and the Rhenish Massif, for example, Tertiary/Quaternary volcanism and degassing of CO 2. Teleseismic traveltime tomography studies show reduced seismic velocity at least down to 250 km depth beneath the western Bohemian Massif, but do not give any clear evidence for the existence of a diapiric low-velocity anomaly which could be interpreted as a mantle plume anticipated beneath the Eger Rift (Plomerová et al. 2007; Koulakov et al. 2009). A regional receiver function study by Geissler et al. (2008) showed delayed arrival times of Ps phases from the 410 km and 660 km discontinuities in central Europe, while the differential traveltimes for the transition zone are as predicted by the Earth reference model IASP91 (Kennett 1991; Kennett & Engdahl 1991). This indicates that the delays are caused mainly by the seismic velocity structure (v p /v s ) in the upper mantle. In this study we show results of a P-receiver function analysis of the discontinuities at 410 and 660 km depth which was aimed

4 Plume search beneath Bohemian Massif 579 Table 1. Station abbreviation, latitude, longitude, altitude, institution that owns the seismic station, data period used in this study, number of receiver function traces, measured delay time of the conversion signals ascribed to the 410 and 660 and delay time difference between P660s and P410s for each of the 117 station sites used in this study. Station Lat ( N) Lon ( E) Alt (m) Institution Data period n tp410s (s) tp660s (s) dt (s) A GFZ/SZGRF A GFZ/SZGRF/Univ. Jena , 09/ / A03/WERN / / /670 GFZ/SZGRF/Univ. Leipzig , , 03/ / A GFZ/SZGRF A GFZ/SZGRF A GFZ/SZGRF A GFZ/SZGRF A GFZ/SZGRF A GFZ/SZGRF A11/BG GFZ/SZGRF/GIPP , 04/ / A GFZ/SZGRF A GFZ/SZGRF A GFZ/SZGRF A GFZ/SZGRF A GFZ/SZGRF A GFZ/SZGRF A GFZ/SZGRF A GFZ/SZGRF A GFZ/SZGRF A21/REGN GFZ/SZGRF/GIPP , 01/ / A GFZ/SZGRF B IG_Prague 07/ / B IG_Prague 09/ / B IG_Prague 04/ / B11/BM EOST/IG_Prague 10/ / B12/BM LITHOSCOPE/IG_Prague 10/ / B15/BM LITHOSCOPE/IG_Prague 10/ / B EOST 10/ / B IG_Prague 02/ / B IG_Prague 05/ / B IG_Prague 03/ / B IRSM 08/ / B IRSM 08/ / B IRSM 08/ / B LITHOSCOPE 08/ / B LITHOSCOPE 08/ / B LITHOSCOPE 08/ / B EOST 03/ / B EOST 03/ / B EOST 03/ / B EOST 03/ / B LITHOSCOPE 08/ /

5 580 B. Heuer et al. Table 1. (Continued.) Station Lat ( N) Lon ( E) Alt (m) Institution Data period n tp410s (s) tp660s (s) dt (s) B LITHOSCOPE 08/ / B LITHOSCOPE 08/ / B LITHOSCOPE 08/ / B LITHOSCOPE 08/ / B LITHOSCOPE 08/ / B LITHOSCOPE 08/ / B LITHOSCOPE 08/ / B LITHOSCOPE 08/ / B LITHOSCOPE 08/ / BAC GFZ/SZGRF BDE Jena 08/ / BG GIPP 02/ / BG GIPP 02/ / BG GIPP 02/ / BG GIPP 03/ / BG GIPP 04/ / BG GIPP 04/ / BG GIPP 04/ / BG GIPP 05/ / BG GIPP 06/ / BG GIPP 06/ / BG Leipzig 02/ / BG GIPP 03/ / BG GIPP 04/ / BG GIPP 05/ / BG GIPP 05/ / BG GIPP 05/ / BG GIPP 06/ / BG GIPP 06/ / BG GIPP 07/ / BG GIPP 04/ / BG GIPP 04/ / BG23/BOH GFZ/SZGRF/GIPP , 04/ / BG GIPP 04/ / BG GIPP 04/ / BG GIPP 04/ / BG GIPP 07/ / BG GIPP 09/ / BG GIPP 07/ / BM IG_Prague 10/ / BOH GFZ/SZGRF BOH GFZ/SZGRF

6 Plume search beneath Bohemian Massif 581 Table 1. (Continued.) Station Lat ( N) Lon ( E) Alt (m) Institution Data period n tp410s (s) tp660s (s) dt (s) BOH GFZ/SZGRF BOH5/BM13/B GFZ/SZGRF/IG_Prague/EOST , 10/ / BRG GRSN 01/ / CLL GRSN 01/ / FALK KTB/Munich/SZGRF 1995, 02/ / GEC GRSN 01/ / GRA SZGRF 01/ / GRB SZGRF 01/ / GRC SZGRF 01/ / GUNZ GFZ/SZGRF/Univ. Leipzig , 02/ / KHC IG_Prague 01/ / KLIN GFZ/SZGRF LAC IG_Prague 05/ / MOX GRSN 01/ / NALB GFZ/SZGRF , 05/ / (24.5) NEUB GIPP 04/ / NIC Jena 07/ / NKC IG_Prague 01/ / NOTT KTB/Munich OTR GIPP 04/ / (43.9) 68.0 (24.1) PRU IG_Prague 01/ / PST Jena 07/ / PVCC IG Prague 09/ / REU Jena 07/ / ROHR Leipzig 04/ / ROTZ KTB/Munich SBG GFZ/SZGRF , 05/ / SELB GFZ/SZGRF TANN Leipzig 02/ / TAU Jena 07/ / WERD Leipzig 06/ / WET GRSN 01/ / ZEU Jena 07/ /

7 582 B. Heuer et al. Figure 3. Common station stack traces of all 117 station. The time axis shows the delay time of the phases with respect to the P onset. Stations are sorted alphabetically. n is the number of individual traces stacked (slowness-corrected). Theoretical arrival times of the converted phases from the 410 and 660 according to the IASP91 reference model are marked by thin vertical lines at 44.1 s and 68.1 s, respectively. The traces are filtered with a bandpass of 1 12 s for the time window between 10 s and 35 s delay time. Later than 35 s delay time, a bandpass of 3 20 s was used. The most prominent signals are the P-to-S conversion at the Moho (PMs) and its multiples (mm) at approximately 11 s to 18 s, as well as the P-to-S conversion at the 410 km discontinuity (P410s) and the 660 km discontinuity (P660s).

8 Plume search beneath Bohemian Massif D Hamburg Berlin 410 km 660 km PL Prague CZ Munich Vienna SK CH A H Figure 4. Piercing points of the P-to-S converted phases at interfaces at 410-km (orange) and 660-km depth (blue). The piercing points are located km and km away from the recording station for the 410- and 660-km discontinuity, respectively. For both depth values the same division of the area into boxes of 1 1 plus a rim of 0.25 (shown only for box 1 by a dashed line) was chosen. Hence, number and location of the piercing points within a specific box are different for interfaces at 410- and 660-km depth. at investigating possible influences of the suggested plume-like structure on the topography of the discontinuities. Our densely spaced network of permanent and temporary stations allows for a much more detailed investigation of the structure of the 410 km and 660 km discontinuity than in the studies by Geissler et al. (2008) and Grunewald et al. (2001). However, our station network has a smaller aperture than the networks in these previous studies. The two main seismic discontinuities in the mantle at 410 and 660 km depth are generally attributed to pressure- and temperatureinduced phase transformations in olivine and other minerals (e.g. Helffrich 2000; Lebedev et al. 2002). Olivine is transformed to a spinel structure at a depth of about 410 km and breaks down to perovskite and magnesiowüstite at about 660 km. These exothermic and endothermic phase transformations have Clapeyron slopes dp/dt of approximately +3MPaK 1 at the 410 and 2 MPa K 1 at the 660, respectively (Helffrich 2000). Thus, if the 410- and 660-km discontinuities are entirely due to these phase transformations, regions of abnormally low temperature such as subduction zones should correspond to elevation of the 410 to lower depths and depression of the 660 to greater depths, so that the distance between the discontinuities is increased. Regions of increased temperature like plume conduits and their surroundings should correspond to depression of the 410 and elevation of the 660, so that the distance between the discontinuities is reduced. Applied to receiver function studies, this means that an increased temperature of K in the mantle transition zone corresponds roughly to a depression of the 410 by about 10 km, which delays the P-to-S converted phase by about 1 s compared to the IASP91 model. As in previous studies, we use the differential delay times of P-to-S converted phases at the 410 and 660 as an estimate for the thickness of the mantle transition zone (thickness according to IASP91 is 24.0 s or 250 km). 2 DATA The study was carried out with data from the international passive seismic experiment BOHEMA (BOhemian Massif HEterogeneity and Anisotropy ) (Plomerová et al. 2003), which includes regional permanent observatories, and data from a previous temporary experiment by Geissler et al. (2005). Thedistributionof the 117 stations is shown in Fig. 1, location and further parameters are given in Table 1. The selected teleseismic events have epicentral distances between 30 and 95, magnitudes larger than 5.5 and a clear P onset with a signal-to-noise ratio larger than 3. The distribution of events is shown in Fig. 2. Altogether, more than 4500 individual receiver function traces were used for investigating the upper-mantle discontinuities. 3 METHOD TheP receiver function method has been used successfully for quite a long time to investigate crustal and upper-mantle structure. The theoretical background of the technique was described, for example, by Vinnik (1977), Langston (1979), Kind & Vinnik (1988), Kosarev et al. (1993) and Zandt et al. (1995). This technique analyses waves that are converted from P to SV waves at a horizontal seismic interface, neglecting seismic anisotropy. The delay time between the

9 584 B. Heuer et al. P410s Delay time [s] P660s n station 30 LAC2 15 BG GRA1 87 BG25 15 A NKC 47 NEUB 54 PVCC 82 ROHR 89 A21/REGN 36 WERD 29 BG16 49 BG10 24 A15 40 B11/BM11 36 BG30 15 OTR 43 BG05 52 BG17 52 BG19 15 BG24 21 KLIN 66 BG02 92 BG23/BOH2 25 A06 15 BAC 111 A11/BG12 20 BOH4 95 A03/WERN 69 B09 61 BOH5/BM13/B13 15 BOH1 55 TANN 201 KHC 139 GEC2 29 FALK 116 BG01 26 GUNZ 28 BG13 68 BG08 46 BG03 43 A BRG 165 WET 172 PRU 23 A12 18 A07 23 NALB 96 BG04 51 BG GRC1 147 GRB1 155 MOX 16 BG20 40 BG29 32 ZEU 173 CLL Figure 5. Sum traces at stations that yielded at least 15 high-quality individual receiver function traces. Explanation see Fig. 3. The traces were sorted by the measured delay time of the P410s conversion signal, marked by red circles. The measured delay times of the P660s conversion signal are marked by green circles. For traces with no circle, no clear conversion signal could be determined. Stations showing no clear signal of P410s are randomly distributed over the investigated area. converted wave and the direct P wave is used as a measure for the depth location of the seismic interface. The method basically comprises the following steps: restitution of the broad-band ground displacement to make signals of different sensors better comparable, component rotation from the ZNE coordinate system into the LQT coordinate system, deconvolution in the time domain as a source-equalization, and finally moveout correction, filtering, and summation of many traces to improve the signal-to-noise ratio of the converted phases. The moveout correction is carried out for a reference epicentral distance of 67 (slowness of 6.4 s deg 1 ). The time of the P onset is set to zero, so that the onset time of all following converted phases is referred to as delay time relative to the P onset. Before the summation of individual traces, their quality was visually checked. Apart from the criterion that the signal-to-noise ratio of the P wave was at least 3, further quality criteria were low noise in the receiver function traces before P onset, existence of a Moho signal and/or Moho multiple reverberations, no extreme noise-like unrealistically large amplitudes or ringing in the later part (20 or more seconds after P onset) of the receiver function trace. Thus, we obtained 4508 individual receiver function traces. We carried out filter tests to maximize the energy of the converted phases at about 44 s (discontinuity at 410 km) and 68 s (discontinuity at 660 km) and decided to use a bandpass filter from 3 s to 20 s. The amplitudes of P-to-S conversions from the upper-mantle discontinuities (named P410s and P660s hereafter) are about four times weaker than the Moho conversion signal (PMs) and not always obvious in single receiver function traces. Hence, it is necessary to stack individual traces to reduce noise and enhance converted phases from the 410 and 660. We used two different approaches to sort the receiver function traces: first, we read delay times of the conversions at 410-km and 660-km depth from common station stacks (Common station method, CSM) with at least 15 individual traces and good signal quality (approach of Geissler et al. 2008, see Fig. 3). In this approach, mainly effects below the stations (in the uppermost mantle) are emphasized. Second, we stacked the data according to their piercing points at 410 km and 660 km depth, respectively (similar to Grunewald et al. 2001). In this approach (CCM), mainly effects at the 410 and 660 are emphasized. Effects from other depth regions superimpose destructively by trend. Due to the ray geometry, the information on the 410 and 660 in the receiver functions originates from a region ca km ( 410 ) and km ( 660 ) away from the receiver. Ps piercing points were calculated using the IASP91 Earth reference model (Kennett 1991). We divided the area into 43 geographical boxes of 1 1, which corresponds to approximately 70 km (E W) 112 km (N S) at the surface (see Fig. 4). We added a rim of 0.25 around each box so that they overlap, which slightly smoothes the results of neighbouring boxes. The size of the resulting boxes corresponds approximately to the dimensions of the Fresnel zone and hence to the resolution of our data. The size of the Fresnel zone for data with a dominant period of 3 5 s is about 200 km at 410 km depth (Grunewald et al. 2001). It is therefore not possible to resolve structures significantly smaller than 150 km laterally at the depth of the mantle transition zone with the receiver function method. As most events were recorded from northeastern directions, the area northeast of the western Bohemian Massif, between Berlin and Prague, is sampled best. More recordings from stations in southeastern Germany would be necessary to get a better resolution of the 410 and 660 beneath the western Bohemian Massif. Nevertheless, there is still fairly good data coverage beneath and to the west of the western Bohemian Massif. To the south, there is a data gap especially in the piercing points at 660 km depth. To study the effect of lithospheric/asthenospheric structure on the delay times of P410s and P660s, we analysed synthetic seismograms in the same way as the real data (see the Appendix). The reflectivity method was used to compute the theoretical seismograms (Kind 1985). To assess the robustness of our data and the influence of noise, we tested different distributions and sizes of the boxes, and stacked different subsets of the data within individual boxes. The observed variations represent mean topographic variations of the mantle discontinuities within the boxes. The error introduced by picking delay time values and by bandpass filtering is relatively small (±0.1 s). Together, this amounts to uncertainties of ±0.3 s for delay time values of both the 410 and 660 obtained with the CCM. For the transition zone thickness, the maximum uncertainty for delay time differences adds up to ±0.6 s. 4 OBSERVATIONS As already mentioned before, we applied two stacking approaches to study the shallow and the deep upper mantle. First, we look at common station stacks, second at common conversion point stacks.

10 Plume search beneath Bohemian Massif 585 Figure 6. (a) Map of the delay time values measured for the P410s conversion signal at each station. One colour comprises 1 s. Green colour indicates a value close to the prediction by the IASP91 reference model (44.1 ± 0.5 s). Red shades indicate increased delay time values, corresponding to a deepening of the discontinuity due to increased temperature, or to decreased seismic velocities above the discontinuity. Blue shades indicate smaller delay times, corresponding to an updoming of the 410 due to relatively lower temperature, or to increased seismic velocities above the discontinuity. For grey coloured stations, no value could be determined from the sum trace. The light grey crosses show the distribution of the piercing points at 410-km depth (see Fig. 4). (b) Like Fig. 6(a), but for the discontinuity at 660-km depth. Note that the colour scale is reversed, because for the 660 the Clapeyron slope of the phase transition is also reversed. That means that increased temperature near the discontinuity would lead to updoming of the discontinuity, while lower temperature would lead to a downshift. Green colour indicates a value close to the prediction by the IASP91 reference model (68.1 ± 0.5 s). (c) Like Fig. 6(a), but for the differential time between P660s and P410s. Green colour indicates normal thickness as predicted by the IASP91 reference model (24.0 ± 0.5 s). Blue shades indicate increased delay time differences, which correspond to increased thickness of the mantle transition zone due to relatively cool temperature. Red shades indicate smaller delay time differences than in IASP91, corresponding to a thinner mantle transition zone due to relatively high temperature. 4.1 Common station stacks This is the usual approach in areas with wide station spacing. However, also in areas with dense station distribution this approach can give important information on the uppermost mantle structure. Individual traces from different azimuths and with piercing points relatively far away from the actual station are stacked, however, the shallow structure below the stations is sampled by rays from all events. Fig. 3 shows the complete sum traces of all 117 station sites analysed. The most prominent signal is the conversion from the Moho (PMs), followed by crustal reverberations (mm). The conversions from the discontinuities at 410 km and 660 km depth can be seen at about 44 s and 68 s, respectively. Unfortunately, not all stations have enough single traces or show reliable Ps phases from the mantle transition zone. Therefore, only the good stations with at least 15 single traces are shown in Fig. 5. For 19 stations we observe clear arrivals from the 660, but not from the 410. This might be due to higher order reverberations from shallow interfaces or complicated structure at the 410. The measured delay time values are marked by red and green circles and listed in Table 1. Table 1 also gives the delay time difference between the two discontinuities as a measure of the thickness of the mantle transition zone. The P410s delay times vary between 43.8 s (BG29) and 46.7 s (BG19) (Table 1). In the corresponding map (Fig. 6a), this is shown by green and reddish colours meaning normal and slightly increased delay times compared to the Earth reference model IASP91. However, some structure is visible: normal values of 44.1 s ± 0.5 s dominate in the N and W of the investigated area, while red shades show increased delay times in the centre and SE (in the Bohemian Massif). The grey stations are stations where we could not determine a delay time value in the sum trace, mostly because of a lack of individual traces, or sometimes because of a missing clear coherent conversion signal. The P660s delay times vary between 67.1 s (CLL) and 69.5 s (LAC2). Similar to the 410, normal values (68.1 s ± 0.5 s according to IASP91 reference model) can be observed in the N and W, while slightly increased delay times dominate in the centre and SE (Fig. 6b). The delay time difference between the two discontinuities

11 586 B. Heuer et al. Figure 6. (Continued.)

12 Plume search beneath Bohemian Massif 587 Figure 7. (a) Data example (box 11, see Fig. 4) in the time window between 30 to 90 s after the P onset. Which data the box actually contains depends on the considered depth (see piercing points in Fig. 4). Left-hand side: traces were sorted according to piercing points at 410-km depth. Right-hand side: traces were sorted according to piercing points at 660-km depth. Single traces were filtered between 3 and 20 s, moveout-corrected and stacked. On top of the single traces, the sum trace is displayed. The P410s and P660s conversion signals are sometimes weak in the single traces but clear in the sum trace. Thin lines at 44.1 s and 68.1 s give delay times predicted by the reference model IASP91. (b) The same for box 19. varies between 21.6 s (BG19) and 24.7 s (GRC1). As visible in Fig. 6(c), normal values around 24.0 s dominate the whole investigated area. However, in the centre and north they are mixed with red shades corresponding to shorter time differences. Since most of the analysed events arrive from northern to eastern backazimuths, the sampled region is slightly shifted to the NE of the individual stations. 4.2 Common conversion point stacks For this stacking approach, we divided the investigated area in 43 geographical boxes as described in Section 3 and shown in Fig. 4. This is possible because we have a dense station spacing and thus a dense coverage of piercing points of the rays at 410 km and 660-km depth. For the analyses we considered only boxes containing at least 20 individual traces. The number of traces within a certain box is usually different for piercing points at 410 km and 660 km depth. Figs 7(a) and (b) show examples of two boxes, displaying all individual traces as well as final common conversion point stacks (sum trace on top). Fig. 8(a) shows the common conversion point stacks for the 410. The measured delay times vary between 43.8 s (Box 2) and 45.2 s (Boxes 20, 31 and 38) (Table 2) with an average of 44.7 s. In the top traces in Fig. 8, only a poor signal of the 410 is recorded. This might again be due to higher order reverberations from shallow interfaces or complicated structure at the 410. The values are shown colour-coded in Fig. 9(a). While the boxes in the north and northeast show normal values in regard to the reference model IASP91 (i.e ± 0.5 s, green colour), most boxes in the centre, west and southeast of the investigated area show delays of conversion signals from the 410. Boxes 27 and 29 seem to be anomalous. However, Table 2 shows that the delay time values in both boxes (44.5 s) is very close to pink values starting at 44.6 s. A similar image is shown in Fig. 8(b) for common conversion point stacks for the 660. Here the delay times of individual boxes vary between 67.9 s (Boxes 14 and 18) and 70.5 s (Box 43, Table 2) with an average of 68.7 s. The values are shown colour-coded in Fig. 9(b). There is a clear structure visible: in the north and partly the centre of the investigated area, delay times of the 660 show normal values of 68.1 ± 0.5 s (green colour), according to IASP91. Southwards, the signals are clearly and increasingly delayed (blue shades) by up to more than 2 s. The time difference between P660s and P410s is also given in Table 2 and plotted into Fig. 9(c). Values vary between 22.9 s (Box 7) and 25.7 (Box 39) with an average of 24.0 s. A clear majority of the boxes shows normal values of 24.0 ± 0.5 s according to IASP91. However, there are also a few boxes showing slightly increased (blue) or decreased (pink) values. Table 2 shows that the blue boxes 2, 8 and 29 have values not far away from normal (green) values. However, between neighbouring boxes 7 and 8 as well as 31 and 40 there remains quite a large step.

13 588 B. Heuer et al. Figure 7. (Continued.) 5 DISCUSSION 5.1 An anomaly within the upper mantle? To answer this question, we look at the observations of the common station method (CSM) as they are more sensitive to shallow structure. We observed roughly similar patterns for conversions at the 410 and 660, that is, normal delay time values in the north and west and increased delay times in the centre and southeast (Figs 6a and b). Comparison with the synthetic data (Fig. A1, Table A1 in the Appendix) shows that the observed delays might be either caused by reduced velocities (increased v p /v s ) in the upper mantle (e.g. asthenosphere material in models BOH-1D and -1E), or by increased temperature of about 200 K at the depth of the 410 (models BOH- 3D and BOH-3I). Obviously, an increased temperature of the whole mantle transition zone does not fit with our data, as this would decrease delay times of the 660 compared to IASP91 (models BOH-3G, -3C and 3D in the Appendix). An increased temperature at the 410 and decreased temperature at the 660 might fit the data but seems unrealistic. Therefore, the most plausible explanation for the observed common station delays is a limited volume of lower velocities in the upper mantle above the 410, which may reach down to the 410 and depress it. This is in accordance with the results of traveltime tomography (Plomerová et al. 2007; Koulakov et al. 2009). The thickness of the mantle transition zone (Fig. 6c) shows values as predicted by IASP91. There are several stations that show decreased differential delay times (reddish colours in Fig. 6c) in the central and northeastern part of the area. This could indicate that a possible temperature anomaly in the upper mantle is reaching down to the 410 km discontinuity, but keeping in mind the uncertainty of ±0.6 s and the size of the corresponding Fresnel zones this interpretation is not supported by the small aperture data.

14 Plume search beneath Bohemian Massif 589 Figure 8. Stacked receiver functions of boxes containing at least 20 traces with piercing points at 410-km depth (a) and 660-km depth (b). The stacked traces are sorted by delay time of P410s and P660s, respectively. The horizontal axis shows the delay time with respect to the P onset. n: number of stacked traces. The single traces were corrected for moveout and filtered between 3 and 20 s before stacking. Thin lines at 44.1 s and 68.1 s mark the delay times of the 410- and 660-km discontinuities according to the IASP91 reference model. Coherent positive arrivals near the delay times predicted by IASP91 are clearly visible for both the 410 and 660. Our measurements of delay times marked by red circles for the 410 and green circles for the 660 (see Table 2). For sum traces with no circle, no clear conversion signal or time pick could be determined. 5.2 An anomaly in the mantle transition zone? In contrast to the CSM, considering the data by common conversion point (CCM) takes the real position of the P-to-S conversion into account, allowing a detailed study at depths of the mantle transition zone. In this approach, traces recorded by different stations are stacked, which means that existing near-receiver structure is smeared over a wide area. For the discontinuities at 410 and 660 km depth, the CCM (Figs 9a and b) generally shows a similar structure as the CSM (Figs 6a and b). Additionally, a deepening of the 660 can be observed in the very south. The differential delay times for the 660 and 410 (Fig. 9c) generally concur with the value predicted by IASP91 reference model (24.0 s). Taking the size of corresponding Fresnel zones and uncertainty of ±0.6 s into account, this indicates normal thickness and temperature of the mantle transition zone. Boxes 18 and 19 might be interpreted as indicative of a thinning of the mantle transition zone. Comparing with the synthetic models, an increased temperature in the upper mantle above (pronounced asthenosphere or plume-like Figure 8. (Continued.) structure) and at the 410 is indicated, possibly combined with a slight depression of the 410 km discontinuity. This means that with the CCM we clearly see a velocity reduction in the upper mantle. However, a shift to north or northeast (our main azimuthal directions) and the depth extend down to the 410 km discontinuity cannot be determined with confidence. Better azimuthal ray coverage is therefore essential for this approach. However, the reduced thickness shown in boxes 18 and 19 (Fig. 9c) can serve as an upper limit for estimating a possible temperature increase at the 410 km discontinuity by about K. 5.3 Implications for the structure of the upper mantle Our findings of a low-velocity anomaly in the upper mantle beneath the northwestern Bohemian Massif and surroundings are based on delays of the Ps conversion signals from the 410 and 660 of up to more than 2 s, compared to IASP91 reference model. We also have some weak signs that this anomaly may extend even down to the discontinuity at 410 km depth. Our results support the results of previous studies. Geissler et al. (2005, 2008) also found delays of up to 2 s at some stations in the same area. However, they used much less data as in this study. In their P-wave tomography study, Piromallo & Morelli (2003) found a pronounced low-velocity

15 590 B. Heuer et al. Figure 9. (a) Delay times measured for the 410-km discontinuity (see Fig. 8a and Table 2). Only boxes containing at least 20 traces were taken into account. Red colours show longer and blue colours shorter delay times than predicted by the IASP91 reference model (44.1 ± 0.5 s, green), which could correspond to relatively hot and cool temperature, respectively. In hatched boxes, the discontinuity was not covered by at least 20 piercing points, or the signal was unclear. The yellow star marks the centre of the Vogtland/northwest Bohemia swarm earthquake region. (b) Delay times measured for the 660-km discontinuity (see Fig. 8b and Table 2). Only boxes containing at least 20 traces were taken into account. The colour scale is opposite to the scale in Fig. 7 as the 660-km discontinuity has a negative Clapeyron slope dp/dt. Blue colours show longer and red colours shorter delay times than predicted by the IASP91 reference model (68.1 ± 0.5 s, green), which could correspond to cool or hot temperatures near the discontinuity, respectively. (c) Difference of the delay times between the P660s and P410s conversions. In the IASP91 reference model, it is predicted to be 24.0 ± 0.5 s. Blue colours indicate a thicker, red colours a thinner mantle transition zone than predicted by IASP91. Coverage with piercing points is not shown because it is different for the two discontinuities within one box (see Fig. 4). anomaly of about 2 per cent in the km depth range beneath the Bohemian Massif as well as beneath the Rhenish Massif and Massif Central. However, it disappears towards the transition zone. Plomerová et al. (2007) found a broad low-velocity anomaly beneath the Eger Rift rather than a plume-like, columnar upwelling, which they interpret as an upwarping of the lithosphere asthenosphere transition. Koulakov et al. (2009) present a model of P- and S-wave velocity anomalies beneath Europe from tomographic inversion. Both P and S models show a low-velocity zone of 2 per cent only in the uppermost mantle between ca. 80 and 250 km depth, yet they deliberately leave the question about the presence of a hypothetical plume at greater depths open. Beneath the geologically comparable Eifel volcanic region, Grunewald et al. (2001) found an apparent depression of the 410 km discontinuity by about 20 km, which points to a zone of increased (by about K) temperature and/or reduced seismic velocities in the mantle above 400-km depth which they interpreted to support the idea of a plume-like structure in the region. Beneath the western Bohemian Massif, their data show normal to slightly increased depth of the 410. However, the results regarding the Bohemian Massif are based on very few stations. No indication for an anomaly of the 660 was found under the Eifel, which argues against a strong thermal anomaly at the bottom of the mantle transition zone (Grunewald et al. 2001; Budweg et al. 2006). The same is the case for the 660 beneath the western Bohemian Massif investigated in our study. However, we need to point out that a structure significantly smaller than 150 km laterally cannot be resolved by the receiver function method. The velocity anomaly in the upper mantle observed by Koulakov et al. (2009) is of the dimension of 150 km. The 1 s isochrones at 410 km depth have a diameter of comparable size (200 km, Grunewald et al., 2001). P-wave tomography by Piromallo & Morelli (2003) does also not show any significant velocity anomaly in the depth range between 600 and 700 km. The tomographic inversion by Koulakov et al. (2009) supports these findings. In the southern part of the investigated area, that is, at the transition between the southern Bohemian Massif and the Eastern Alps, we map a clear deepening of the 660 in the CCM. The delay of arrival times of up to more than 2 s corresponds to a deepening of the 660 of about 20 km or a decreased temperature of about K. However, this southernmost area is in the geotectonic context of the Alpine foreland and beyond the scope of our study The thickness of the mantle transition zone is normal in most parts of our investigated area. There is a hint to a slightly thinned mantle transition zone in boxes 18 and 19. We consider the increased thickness of the mantle transition zone in boxes 39 and

16 Plume search beneath Bohemian Massif 591 Figure 9. (Continued.)

17 592 B. Heuer et al. Table 2. Box number, number of stacked traces for each box for the 410 and 660, delay times of the Ps conversions t P410s and t P660s and delay time difference for the conversions from the two discontinuities within each box. Box number n (410) n (660) t P410s (s) t P660s (s) dt (s) as significant (Fig. 9c), even though it is at the rim of the investigated area. It is in accordance with the results of Geissler et al. (2008) and Piromallo & Morelli (2003). Koulakov et al. (2009) show increased P velocities at a depth of 500 km in the very south of Germany, while at 700 km depth the velocity structure is not anomalous. If the manifold processes that point to magmatic activity beneath the western Bohemian Massif should have their origin in a diapiric mantle upwelling as predicted by Granet et al. (1995), the absence of a distinct topography of the upper-mantle discontinuities leads to the conclusion that this predicted mantle finger shows only a weak imprint on the 410 km discontinuity and no effect on the 660. If it exists, it either needs to have its origin above the mantle transition zone, or its diameter is significantly smaller than 150 km at the depth of the mantle transition zone and thus too narrow to be resolved by teleseismic receiver functions. This is in accordance with the tomographic study of Koulakov et al As already stated in Geissler et al. (2008) and references therein, the synthetic analysis shows that most of the apparent deepening/upwelling of the upper-mantle discontinuities can be explained by reduced/increased seismic velocities in the crust and uppermost mantle. This was already proposed by Kind & Vinnik (1988). The necessary higher v p /v s ratios can be partly caused by different chemical composition, as well as by anisotropy. 6 CONCLUSIONS Consistent delays of the Ps converted waves from the 410 and 660 in common station stacks clearly indicate a low-velocity anomaly (high v p /v s anomaly) in the upper mantle beneath the northwestern Bohemian Massif and surroundings, in accordance with previous studies. This is supported by delays of the Ps converted waves from the 410 and 660 in common conversion point stacks. There are slight indications that this anomaly may extend down to the discontinuity at 410-km depth, because decreased differential delay times between conversions from the 410 and 660 were observed by common conversion point stacking. As the IASP91 global reference model does not contain an asthenosphere, an increased v p /v s ratio within an existing asthenosphere beneath the investigation area might explain at least some part of the observed time delay of the discontinuities of the mantle transition zone. Similarly, a plume-like structure in the upper mantle above 410 km may explain the observed data. A deepening of the 660 was observed beneath the southern part of the investigated area by the CCM. The thickness of the mantle transition zone is normal to slightly thinned in the northern and central part of the investigated area, while it increases in the very south. The baby plume beneath the western Bohemian Massif predicted by Granet et al. (1995) would thus have no or only weak impact on the topography of the 410- and 660-km discontinuity. Our conclusion is that also with a considerably increased data set no indication of a plume in the mantle transition zone is detected, however from the receiver function point of view a plume-like structure may exist in the upper mantle below western Bohemia. ACKNOWLEDGMENTS We would like to thank the Geophysical Instrument Pool Potsdam of GFZ Potsdam for providing mobile seismic stations for the BOHEMA experiment. W. Hanka helped with archiving the data in the GEOFON data centre. We are also grateful to all institutes providing data from the permanent observatories. For data processing we used K. Stammler s program SeismicHandler. Most figures were generated using the Generic Mapping Tools (GMT). We would like to thank the reviewers for their helpful and constructive comments. We thank the Deutsche Forschungsgemeinschaft (DFG) for the funding of the BOHEMA experiment, the EU for support within the NERIES project and GFZ Potsdam for further funding of B.H. REFERENCES Babuška, V. & Plomerová, J., Subcrustal lithosphere around the Saxothuringian-Moldanubian Suture Zone: a model derived from anisotropy of seismic wave velocities, Tectonophysics, 332, Bina, C.R. & Helffrich, G., Phase transition Clapeyron slopes and transition zone seismic discontinuity topography, J. geophys. Res., 99(B8), Bräuer, K., Kämpf, H., Niedermann, S. & Strauch, G., Evidence for ascending upper mantle-derived melt beneath the Cheb basin, central