Seismic lithosphere–asthenosphere boundary beneath the Baltic Shield

Abstract

The problem of the existence of the asthenosphere for old Precambrian cratons is still discussed. In order to study the seismic lithosphere–asthenosphere boundary (LAB) beneath the Baltic Shield, we used records of nine local earthquakes with magnitudes ranging from 2.7 to 5.9. To model the LAB, original data were corrected for topography and Moho depth using a reference model with a 46-km-thick crust. For two northern events at Spitsbergen and Novaya Zemlya, we observe a low-velocity layer, 60–70-km-thick asthenosphere, and the LAB beneath Barents Sea was found at depth of c. 200 km. Sections for other events show continuous first arrivals of P-waves with no evidence for “shadow zone” in the whole range of registration, which could either be interpreted as the absence of the asthenosphere beneath the central part of the Baltic Shield, or that the LAB in this area occurs deeper (>200 km). The relatively thin low-velocity layer found beneath southern Sweden, 15 km below the Moho, could be interpreted as small-scale lithospheric heterogeneities, rather than asthenosphere. Differentiation of the lower lithosphere velocities beneath the Baltic Shield could be interpreted as regional heterogeneity or as anisotropy of the Baltic Shield lithosphere, with high velocities approximately in the east–west direction, and slow velocities approximately in the south–north direction.

Keywords::

• cratons
• East European platform
• crust
• lithosphere
• seismic lithosphere–asthenosphere boundary

Introduction

The moving lithospheric plates are crucial elements in global plate tectonics. The lithosphere (from the Greek: λίθ°ς – rock) is understood as an outer shell of the Earth, which consists of the crust and a certain part of the upper mantle. Rigid lithosphere, sometimes called the “lid”, is underlain by ductile asthenosphere (from the Greek: ασθϵνώς – weakly).

Petrological studies of xenoliths and thermal modelling based on heat-flow measurements allow determination of the temperature gradients in the lower lithosphere, give an idea on the composition of the lithosphere and place constraints on the depth of the transitional zone between the lithosphere and the asthenosphere. The lithosphere–asthenosphere boundary (LAB) is not a sharp discontinuity, but rather a gradual and wide transition zone (see e.g. Meissner 1986). From the point of view of elastic properties and seismic velocities, the asthenosphere was identified as a low-velocity channel in the 50–200 km depth range in Gutenberg's (1959) global model of the Earth. Low seismic velocities, or lowering of the velocity gradients with depth, in the upper mantle have also been found in regional scales beneath continents and oceans (e.g. Anderson & Toksöz 1963; Johnson 1967; Schubert et al. 1976; Grad 1988; Bruneton et al. 2004; Rychert et al. 2005, 2010; Gorman et al. 2006; Rondenay et al. 2009). The LAB depth strongly depends on temperature; the thicker lithosphere up to c. 200 km is clearly observed under “cold” Precambrian shields and platforms, while the thinnest lithosphere, of 50–100 km, is found under “hot” oceans and oceanic and continental rifts. Recently, the LAB depth have been effectively estimated from velocity variations in global and regional tomography models, group velocities of long period (c. 100 s) surface waves and S-wave receiver functions (e.g. Bruneton et al. 2004; Gregersen et al. 2006; Li et al. 2007; Pasyanos 2010; Wilde-Piórko et al. 2010). The LAB may also correlate with a downward extinction of seismic anisotropy or a change in the anisotropy direction (e.g. Eaton et al. 2009; Plomerová & Babuška 2010). The asthenosphere could be identified as a low-viscosity zone and a low value of the quality factor, Q_S (e.g. Stacey 1969; Karato 2010). However, the composition of the mantle in the pressure–temperature space could also affect on the low-velocity zone. For example, 3D modelling of S-wave velocity beneath the Baltic Shield showed significant lateral heterogeneity across the region, interpreted to arise from compositional variations within the lithospheric mantle (Bruneton et al. 2004). Characteristic for asthenosphere is low seismic activity or even total lack of earthquakes.

The electrically defined LAB is marked by a significant reduction in electrical resistivity, where resistive lithosphere is overlying the highly conductive asthenosphere. The asthenosphere, identified as a high conductivity layer (low resistivity layer) in the upper mantle, often coincides with a seismic low-velocity zone (e.g. Martinec & Wolf 2005; Korja 2007; Eaton et al. 2009).

The lithosphere may also be defined thermally. The use of borehole heat-flow measurements allows calculation of the lithospheric geotherms and estimation of the thermal lithosphere thickness, defined as the depth where the continental geotherms intersect the 1300°C mantle adiabat (e.g. Artemieva & Mooney 2001; Artemieva 2007).

The upper and lower boundaries of the asthenosphere appear to be gradient zones rather than first-order discontinuities. This suggests that we should not expect large contrasts of elastic parameters that would produce strong reflected and converted phases, and that seismic reflection and receiver function techniques may not be optimal in this context. Instead, surface wave methods, seismic tomography and searching for “shadow zones” of P and S body waves appear to be the most promising methods for the determination of the seismic LAB depth.

In this paper, we are using a relatively conventional approach – the “refraction” method, which permits the determination of layer thicknesses (lithosphere) and velocities in order to resolve the depth of the LAB. For a LAB depth of 100–200 km, “shadow zones” are expected at epicentral distances of about 1500–2000 km. For studying the “shadow zones”, local seismic events with relatively big magnitudes, which produce good quality records at long distances from the source, can be used (Fig. 1).

Fig. 1 The seismicity map of central and northern Europe (January 2008 – September 2011) showing locations of 2698 events: 2285 from the Helsinki catalogue (Uski & Raime 2010) and 413 from GEOFON. Events are mostly shallow: 90% of the events have depths < 20 km, maximum depth is 70 km for an event in Spitsbergen; the highest magnitude for an event is 5.9.

The Baltic Shield and previous seismic investigations of the lithosphere and the LAB

The East European Craton (EEC) is the coherent Precambrian part of Europe, mainly of Archaean and Palaeoproterozoic ages, assembled in the Late Palaeoproterozoic (Bogdanova et al. 2005, 2006). In the southwest, the EEC is limited by accreted Phanerozoic terranes of the Trans-European suture zone (TESZ), which extends from the British Isles to the Black Sea region. The northeastern part of the European continent occupies the Baltic (also Fennoscandian) Shield with the Precambrian bedrock exposed on the Earth's surface (Figs. 1 and 2).

The typical three-layer crust of the Baltic Shield has a thickness in the range of 42–60 km and attains its maximum thickness in Central Finland (Grad & Luosto 1987; Luosto 1997). The P-wave velocities in the upper crust are 6.0–6.4 km s^− 1, in the middle crust 6.6–6.9 km s^− 1 and in the lower crust >7.0 km s^− 1. The thickness of upper and middle crusts is in the range of 32–42 km. The thickness of the lowermost high-velocity crustal layer is in the range of 4–24 km. The P-wave velocity in the uppermost mantle is 7.9–8.2 km s^− 1. The ratio of P-wave velocity to S-wave velocity varies typically from 1.67–1.70 in the upper crust to 1.77–1.80 in the lower crust. In the upper mantle, the ratio is c. 1.73 (e.g. Guggisberg 1986; Grad & Luosto 1987; Luosto 1997; Hyvonen et al. 2007).

Similar to other old Precambrian cratons, features of the asthenosphere beneath the Baltic Shield are still discussed (e.g. Artemieva 2003; Bruneton et al. 2004; Kozlovskaya et al. 2008). The first determinations of the lithospheric thickness beneath the Baltic Shield were obtained from analyses of fundamental-mode and higher order Rayleigh surface waves. Dispersion of the higher mode data has been interpreted by Nolet (1977) to distinguish the thick lithosphere of the Baltic Shield from the thinner Western European lithosphere (see also Zielhuis & Nolet 1994). With the use of higher modes, Cara et al. (1980) found no need for a low-velocity zone in the mantle beneath the northern Eurasia. They also argue that a nearly constant 4.5–4.6 km s^− 1 S-wave velocity is required in the uppermost 200 km. Using Rayleigh-wave dispersion data for the Fennoscandian region, Calcagnile (1982) found lid thicknesses of up to c. 135 km in the area from Bothnia to north-central Finland with a weak, if any, shear velocity contrast to the underlying layer. The surrounding areas are characterized by lid thicknesses of up to c. 75 km only. A stronger low-velocity zone with a lid contrast of 0.25–0.45 km s^− 1 may be found in the Caledonian and the Baltic Sea area (Calcagnile 1982). An updated map of the lithosphere–asthenosphere system in the Europe (Panza 1985; Calcagnile & Panza 1987) shows much larger lithospheric thicknesses of the Baltic Shield, in the range of 110–170 km, increasing to >190 km in its central part. The Baltic Shield model obtained later by Dost (1990) shows an absence of the low-velocity layer, while density seems to be lower at 200–350 km depth.

The deep structure of the lower lithosphere and asthenosphere can also be investigated from studies of P-waves. From the explosive source refraction profile FENNOLORA, Guggisberg (1986) interpreted several deep low-velocity channels (with about 5% decrease in P-wave velocity in a few tens of kilometres thick layers) in the lower lithosphere, with the bottom of the lithosphere at about 200 km depth marked by a velocity change of a couple of per cent. Regional 1D S-wave velocity models for the central part of the Baltic Shield, obtained from the SVEKALAPKO array (Kozlovskaya et al. 2008), show that velocities beneath the craton are significantly higher (about 4% down to a depth of 250 km) compared with standard model iasp91 (Kennett & Engdahl 1991). On the other hand, lowering of the velocity with depth is not visible (Kozlovskaya et al. 2008). This fact could be interpreted as an absence of asthenosphere, or that the LAB in this area occurs deeper. Plomerová & Babuška (2010) define the LAB as a boundary between fossil anisotropy in the lithospheric mantle and underlying seismic anisotropy related to present-day flow in the asthenosphere. The LAB topography is more distinct beneath the Phanerozoic part of Europe than beneath its Precambrian part. Beneath the central Fennoscandia, the LAB deepens down to c. 220 km (Plomerová & Babuška 2010).

Seismic data for the Baltic Shield

Central and Northern Europe are characterized by weak–to-moderate seismicity (e.g. Slunga 1991; Ahjos & Uski 1992). The map in Fig. 1 shows locations of 2698 events recorded from January 2008 to September 2011 in the Baltic Shield and its surroundings. Most of these events were classified as earthquakes. Events recorded at less than three stations, as well as explosions from known sites, were omitted from the data-set. Mining-induced seismic events (e.g. rock bursts, mine collapses) were also excluded. However, it is likely that some man-made events still exist in the data, especially among low-magnitude events (Uski & Raime 2010). It should be mentioned that the actual distribution and concentration of events is based on the properties of the network deployed to detect them, particularly in the case of small events.

Fig. 1 shows that seismic events are concentrated in a few distinct areas. In the north (18–20°E, 77–78°N), a group of events is concentrated in the southern part of Spitsbergen, within continental crust of the Barents Sea block (see also Figs. 2 and 3). West of 10°E and north of 73°N, a linear group of epicentres is related to the Knipovich Ridge and Mohns Ridge segments of the Mid-Atlantic Ridge. Events south of this group are related to continent–ocean transition (COT) between oceanic crust of the Atlantic and continental crust of the Barents Sea. The coastal area of Scandinavia is characterized by moderate seismicity of the Caledonides. In the southern part of the study area (6–8°E, 50–52°N), a group of events is related to seismic activity of young continental rifting – in the Rhine Graben. Events occurring around 16°E and 52°N are related to copper mining activity in Lubin in southwestern Poland. There are both natural and induced earthquakes in this area. Events in the area around 19°E and 50°N relate to coal mining activity in Silesia (Draber et al. 2002).

Fig. 2 The location of nine events (numbered black stars) and seismic stations (dots) used in this study on the topography/bathymetry map of the region (Amante & Eakins 2009). The small star with number 10 shows the location of a small event which occurred in western Finland and was recorded together with event no. 9; note that the location of event no. 8 is outside the map frame; for more details, see Table 1. The area of the Baltic Shield is hatched in red and thick red dashed line shows the location of the Ladoga–Bothian Bay shear zone between the Proterozoic Svecofennian (SF) domain in the south and the Archean Karelian (AR) domain in the north. The main tectonic elements of the region are as follows: TESZ, Trans-European suture zone; SW margin of the East European platform; VTE, Variscan terranes of Europe; KR and MR, Knipovich and Mohns ridges, parts of Mid-Atlantic Ridge; COT, continent–ocean transition.

Fig. 3 Moho depth map of the European Plate (Grad et al. 2009). Also shown are the seismic refraction profiles (white lines) with the positions (white dots) where velocities were sampled to compile the average P-wave velocity model of the crust and uppermost mantle of the Baltic Shield. The yellow frame shows the area for which an average Moho depth (45.7 km) was calculated. Data were compiled from Hirschleber et al. (1975), Lund (1979), Guggisberg (1986), Luosto (1986), Grad & Luosto (1987), EUGENO-S Working Group (1988), Luosto et al. (1989, 1990, 1994), Grad et al. (1991), FENNIA Working Group (1998), Uski et al. (2012) and Tiira et al. (2014).

The Baltic Shield in general is a stable intraplate region characterized by weak-to-moderate seismicity, with event magnitudes rarely exceeding 4.0. The majority of earthquakes occurs within the Kuusamo–Kandalaksha region in the northeastern Finland and adjacent Russia (28–35°E, 65–68°N), and along a broad north–south trending belt running parallel to an ancient plate boundary from the Bothnian Bay to northern Norway (from 18°E, 63°N to 25°E, 70°N; Ahjos & Uski 1992). The Kuusamo area is the area where natural earthquakes are the most frequently recorded in Finland. However, they are rather weak, and most of them are detected only because of the local seismic array in this area (Fig. 2). In the southern part of the Baltic Shield, an area of low seismicity in Skåne (11–14°E, 55–57°N) has some relatively strong (with magnitudes up to 5) events. In the Fennoscandian area, 80–90% of all earthquakes occur in the uppermost 20 km of the Earth's crust (Slunga 1991; Ahjos & Uski 1992). Earthquakes at depths of 40 km and deeper are rare. However, some well-determined events in northern Sweden have been located close to the crust–mantle boundary (Arvidsson & Kulhanek 1994; Arvidsson 1996). These observations suggest that the Archaean lower crust may be seismically active and involved in brittle deformation (Uski et al. 2012).

To study the lithosphere–asthenosphere system of Fennoscandia, we need relatively powerful sources (earthquakes or quarry blasts) generating seismic waves that can be recorded up to a distance of at least 2000 km. In the first step of event selection, we chose the strongest events that occurred within last years. We chose to use fairly recent events since this will enable us to use data recorded by stations of the Swedish National Seismological Network (SNSN), a dense regional seismic network of broadband stations allowing good quality records of local events (e.g. Olsson 2007). The number of the SNSN stations has grown rapidly in the last decade, and today the network has a good coverage of most of Sweden. Because of the size of the Baltic Shield (about 1500 km in diameter), we limited our data-set to events from the outskirts of the shield in order to obtain large enough station-to-event distance range. The nine events chosen are listed in Table 1 and shown in Fig. 2 (event no. 8 at Novaya Zemlya is located outside the map frame). The magnitudes of the events are in the range of 2.7–5.9. Exception is the small event no. 10 (magnitude 1.1) in western Finland, which occurred nearly at the same time as event no. 9 in the Rhine Graben (both events were recorded simultaneously). In Fig. 2, earthquakes are marked by black stars with numbers and the seismic stations providing the data are marked by black dots.

Table 1 Events used for lithospheric structure studies in the Baltic Shield area.

Event no.	Date (day month year)	Time (hour:minute:second)	Lat φ (°N)	Long λ (°E)	Depth (km)	M	Place
1	24 January 2008	18:51:51.4	57.63	7.37	10.0	2.7	Skagerak
2	21 February 2008	02:46:17.6	77.01	19.01	15.4	5.9	Spitsbergen
3	13 June 2008	04:13:30.4	51.54	16.07	4.1	4.3	Lubin
4	14 October 2008	16:47:42.2	69.74	30.14	2.7	3.0	Varangerfjord
5	12 December 2008	05:20:03.1	55.57	13.53	19.6	4.9	Skåne
6	25 May 2009	08:05:50.8	67.65	33.99	1.0	3.4	Kola
7	22 January 2010	04:05:42.1	51.18	19.09	4.1	4.3	Bełchatów
8	11 October 2010	22:48:28.0	76.28	63.95	10.0	4.7	Novaya Zemlya
9	14 February 2011	12:43:11.0	50.37	07.79	12.0	4.0	Rhine Graben
10	14 February 2011	12:47:28.7	64.12	24.27	0.0	1.1	Western Finland

Note: Data compiled from Helsinki catalogue, Uski & Raime (2010), Pirli et al. (2010) and Wiejacz & Rudziński (2010).

Average velocities, reference model and time corrections

To determine the seismic velocity structure of the lithosphere and the LAB depth beneath the Baltic Shield, we studied travel times of local earthquakes. The relatively large number of permanent seismic stations in the Baltic Shield permits for relatively dense recording of the events, with 30–100 km distance between the stations. This spacing is sufficient in the LAB studies as the penetration depth of body waves is a few hundreds of km and the wave length is in the order of a few kilometres. Recording stations are not linearly aligned, which would have permitted for easy 2D interpretation of the structures along the profile, but scattered in a wide corridor of a few hundreds of kilometres width. This means that two stations at similar epicentral distance may lie at locations with significantly different altitudes and Moho depths. To model the lower lithosphere and asthenosphere, the original data need to be time-corrected for topography and Moho depth variation for each event and each station location. Topography for seismic station locations changes from close to sea level up to 630 m (see Fig. 2).

Differences in the Moho depth in the study area are significant. As seen from Fig. 3, the Moho depth in the Baltic Shield is varying of 30–60 km (Grad et al. 2009). An average Moho depth was calculated for the area within the yellow frame shown in Fig. 3, using data from the digital Moho depth map by Grad et al. (2009). The average depth is 45.7 km, and in the reference model a value of 46 km depth was used. P-wave velocities in the reference model were compiled from seismic studies in the area. In Fig. 3, the investigated profiles are shown by white lines (Hirschleber et al. 1975; Lund 1979; Guggisberg 1986; Luosto 1986; Grad & Luosto 1987; Luosto et al. 1989, 1990, 1994; Grad et al. 1991; FENNIA Working Group 1998; Uski et al. 2012; Tiira et al. 2014) and places of sampled velocities are shown by white dots. All velocity–depth relations were sampled at 1-km depth intervals. The velocity values were weighted according to reliability of the models. The highest weight was given to data from modern refraction and wide-angle reflection profiles with a dense system of observations and good reciprocal coverage, e.g. SVEKA (Grad & Luosto 1987), BALTIC (Luosto et al. 1990) and POLAR (Luosto et al. 1989). Velocity in the crust and uppermost mantle was directly extracted from 2D numerical models with a spacing of 60 km along profiles. Only parts of models sufficiently sampled by rays were used in velocity–depth analysis. Published 1D models were adopted for older profiles with a sparse system of observations, e.g. Blue Road (Lund 1979), FENNOLORA (Guggisberg 1986) and Sylen-Porvoo (Luosto 1986).

P-wave velocities in the crust and uppermost mantle of the Baltic Shield were sampled in 67 locations (white dots in Fig. 3). Comparison with location of seismic stations (Fig. 2) shows a good coverage for the entire Baltic Shield. In total, 4269 values of velocity were used for the determination of average velocity–depth relations in the crust and uppermost mantle. The data were fitted by linear functions:

(1)

(2)

where V is the P-wave velocity in km s^− 1 and z is the depth in km. The data distribution is shown in grey scale in Fig. 4(a). Apart from fitted relations (1) and (2), the reference model is shown by thick black-dotted line (Fig. 4(a)). The reference model has a 46-km-thick crust with velocities increasing from 5.97 at the Earth's surface to 7.24 km s^− 1 at the 46-km depth. The velocity below the Moho at 46 km depth is 8.13 km s^− 1.

Fig. 4 A. Average P-wave velocities of the crust and uppermost mantle of the Baltic Shield fitted by the linear functions (1) and (2) given in the text. In total, 4269 values of velocity in 67 locations were used (see white dots in Fig. 3). The grey scale shows the density of data in 0.1 km s^− 1 1 × 1 km cells. The reference model shown by the black-dotted line has 46-km-thick crust with velocities increasing from 5.97 at the Earth's surface to 7.24 km s^− 1 at the 46-km depth. The velocity below the Moho at 46 km depth is 8.13 km s^− 1. B. A comparison of travel time curves for the reference model (black line for Moho depth 46 km) with travel time curves for shallower Moho (34 km depth, blue line) and deeper Moho (58 km depth, red line). Reduction velocity is 8 km s^− 1. For more explanation see text.

Travel times for the reference model are shown by the black line in Fig. 4(b). For comparison, two travel time curves for shallower Moho (34 km depth, blue line) and deeper Moho (58 km depth, red line) are shown. Time differences are about ± 2 s at distances between 300 and 500 km, decreasing to about ± 1 s at distances of 1000 to 1500 km. Travel times were calculated with 1-km intervals for Moho depths from 34 to 58 km. The time corrections were calculated using interpolation for both distances and the Moho depths below the event and station. In the modelling of each event, we used “true” depth of the event, making correction only for the Moho depth (difference between the depth of the event and the Moho depth). The final time corrections included also the correction for topography.

An example of the application of time corrections is shown in Fig. 5. The map in Fig. 5(a) shows the location of event no. 3 (Lubin; marked by black star) and seismic stations (black dots) with their numbers. Corresponding sections with and without corrections are shown in Fig. 5(b),(c), respectively. Red dots are picks of first arrivals (the same picks are shown in both sections). Red lines are travel times and grey bands show scattering of the data. Application of the time corrections improves correlation. The scattering of first arrivals is about 1 s for the section with the time corrections (Fig. 5(b)), and about 2 s for the section without corrections (Fig. 5(c)). The histogram in Fig. 5(d) shows the distribution of 604 total time corrections. About 93% of the corrections are in the range from − 0.1 to 0.8 s, with a maximum for the 0.2–0.5 s interval (282 corrections). Such a processing sequence was applied for all sections in this paper, which permits us to use a reference crustal model without topography (0 m above sea level) and with the reference Moho at 46 km depth. After applying the corrections, all sections correspond to Moho depth of 46 km and stations at sea level.

Fig. 5 Example of application of the time correction discussed in the text. A. Map showing the location of event no. 3 (Lubin) marked by star and seismic stations in southern Sweden (numbered dots) used in this comparison. B. Recorded seismograms with time corrections applied. Note much lower scattering of the first arrival times in the corrected section compared to uncorrected section C. C. Recorded seismograms without the time corrections. Red dots are picked first arrivals for P-waves, and the thick grey line shows a scattering range of first arrival times; reduction velocity is 8 km s^− 1. D. Histogram showing the distribution of 604 total time corrections. About 93% of the corrections range from − 0.1 to 0.8 s, with a maximum of 0.2–0.5 s interval (282 corrections).

Searching for asthenosphere beneath the Baltic Shield

To study the LAB in the Baltic Shield, we chose nine events (Table 1, Fig. 2) from the outskirts of the shield. For all record sections, time corrections were applied as described in the previous section. Poor-quality and noisy seismograms were removed from the sections. Some seismograms were also omitted to avoid crowding of seismograms from stations with similar epicentral distances. All record sections were normalized, filtered and drawn with a reduction velocity of 8 km s^− 1 (in some cases with a reduction velocity of 4.5 km s^− 1 for sections of S-waves).

Modelling of all sections in this paper was done using the ray-tracing technique and software SEIS83 (Červený & Pšenčík 1984). As a reference crustal model for P-wave velocity, we used the model described in the previous section. Crustal S-wave velocities were recalculated from P velocities using the relation V_P/V_S = 1.67 in the uppermost crystalline basement and V_P/V_S = 1.77 in the lower crust (see e.g. Grad & Luosto 1987; Bogdanova et al. 2006; Uski et al. 2012). The models of the mantle structure were derived by trial-and-error forward modelling. Travel times and synthetic seismograms were calculated for 1D models with Earth-flattening transformation for waves from a point source (Hill 1972). This transformation permits to calculate travel times for a flat model and transform for a spherical Earth.

Two examples of the record sections for event no. 3 (Lubin) and for event no. 6 (Kola) are shown in Figs. 6 and 7. Apart from the full wave fields of P, S and surface waves (Figs. 6(a) and 7(a)), enlarged parts of the sections are shown for S-waves (Figs. 6(b) and 7(b)) and P-waves (Figs. 6(c) and 7(c)). Red dots are the picked first arrivals of P- and S-waves, and red lines are the first arrivals of P- and S-waves calculated for the iasp91 model (Kennett & Engdahl 1991). Although the crustal thickness in the iasp91 model is only 35 km (in comparison with 46 km for the Baltic Shield), observed P and S arrivals are both significantly earlier than predicted by the iasp91 model – at a distance of 1700 km c. 3–4 s for P-waves and c. 6–8 s for S-waves. This is a clear indication that lithospheric P and S velocities beneath the Baltic Shield are higher than those of the iasp91 model. Also, the observed continuation of the first arrivals up to 1700–2000 km distance could be interpreted as a lack of a “shadow zone”. High lithospheric velocities for the EEC using Pn and Sn waves recorded at teleseismic distances were described nearly 50 years ago by Båth (1966).

Fig. 6 Examples of record sections with time-corrected seismograms for event no. 3 (Lubin). A. Full wave field of P, S and surface waves. Filtration 1–7 Hz, normalized traces, reduction velocity 8 km s^− 1. Red lines are first arrivals of P- and S-waves calculated for the iasp91 model. Enlarged parts of this section are shown for P- and S-waves. B. Section for S-waves, filtration 2–7 Hz, normalized traces, reduction velocity 4.5 km s^− 1. C. Section for P-waves, filtration 2–10 Hz, normalized traces, reduction velocity 8 km s^− 1. Red dots are the picked first arrivals for P- and S-waves, and red lines are the first arrivals of P- and S-waves calculated for the iasp91 model. Phases related to the “410” and “660”-km boundaries are also shown in the sections. Note the early arrivals of both P- and S-waves compared with the iasp91 model – at distance 1700 km about 4 s for P-waves and about 8 s for S-waves.

Fig. 7 Examples of record sections with time-corrected seismograms for event no. 6 (Kola). A. Full wave field of P, S and surface waves. Filtration 1–7 Hz, normalized traces, reduction velocity 8 km s^− 1. Red lines are the first arrivals of P- and S-waves calculated for the iasp91 model. Enlarged parts of this section are shown for P- and S-waves. B. Section for S-waves, filtration 2–7 Hz, normalized traces, reduction velocity 4.5 km s^− 1. C. Section for P-waves, filtration 2–10 Hz, normalized traces, reduction velocity 8 km s^− 1. Red dots are the picked first arrivals for P- and S-waves, and red lines are the first arrivals of P- and S-waves calculated for the iasp91 model. Phases from “410-” and “660”-km boundaries are also shown. Note the early arrivals of both P- and S-waves compared with iasp91 model – at distance 1700 km about 3 s for P-waves and about 6 s for S-waves.

Sections with records up to 2200–3000 km distances are shown in Figs. 8 and 9. For event no. 8 (Novaya Zemlya), good-quality first arrivals of P-waves are recorded in the distance range of 1300–3000 km. At an epicentral distance of about 1800 km, a clear “shadow zone” is visible (marked in Fig. 8 by an arrow). Corresponding rays are shown in the ray diagram: red rays travelling in the lithosphere and navy blue rays reflected from the bottom of the asthenosphere and mantle including “410-” and “660”-km boundaries.

Fig. 8 Record section for event no. 8 (Novaya Zemlya). Filtration 1–5 Hz, normalized traces, reduction velocity 8 km s^− 1. The beginning of the “shadow zone” of first arrivals is observed at a distance of c. 1900 km (shown by arrow), and the corresponding depth of the low-velocity zone (LAB) is 201 km. Red dots are picked first arrivals of P-waves, and red lines are travel times. The V_P model (black line) is shown together with the iasp91 model (yellow line). The bottom left plot shows rays in the mantle (red in the lithosphere and navy blue in deeper mantle).

Fig. 9 Record sections of two events (with and without asthenosphere), together with travel times of the best fitted models. A. Section of event no. 2 (Spitsbegen). Filtration 2–15 Hz, normalized traces, reduction velocity 8 km s^− 1. The beginning of the “shadow zone” of first arrivals (shown by arrow) is observed at a distance of c. 1800 km, and corresponds to a depth of the low-velocity zone (LAB) at 198 km. B. Synthetic seismograms for event no. 2 (Spitsbegen). For synthetics, a complex source was simulated with a length of radiation of around 20 s. The figure shows the difficulties in interpretation of secondary arrivals masked by a long-source function. However, the “shadow zone” in first arrivals is still clear, which permits conclusions about the LVL – asthenosphere. C. Section for event no. 7 (Bełchatów). Filtration 2–15 Hz, normalized traces, reduction velocity 8 km s^− 1. A continuity of the first arrivals is observed in the whole distance range up to 2000 km. In this case, the low-velocity zone (asthenosphere) must be deeper than 180–200 km, if it exists at all. In all sections, red dots are picked first arrivals of P-waves, and red lines are fitted travel times. For event no. 7 (Bełchatów), a test of resolution of calculated travel times was done. The arrival times of P phase with the velocity in the lower lithosphere lid perturbed by +0.05 and − 0.05 km s^− 1 shown by blue and orange lines are significantly too early and too late relative to the final model (see red curve in Fig. 14(b)). This indicates that the average lid velocity could be determined with an accuracy better than ± 0.05 km s^− 1.

Also shown in Fig. 8 is a comparison between the 1D velocity model for this event and the iasp91 model. An interesting observation is related to the frequency of the first arrivals. The lithospheric phases have significantly higher frequencies than the deeper phases and this change is observed at relatively short epicentral distance (compare section at distances 1900 and 2100 km). One explanation could be strong attenuation of high-frequency content for a pulse travelling through the asthenosphere. In such a case, the asthenosphere should be characterized by a lower value of the quality factor Q_P. A broad zone of low velocities and low values of Q_P was found beneath the FENNOLORA profile between depths of 100 and 160 km (Hauser & Stangl 1990; Stangl 1990).

For event no. 2 (Spitsbergen), good-quality P-waves are recorded in the distance range of 800–2400 km (Fig. 9). First arrivals are far away from explosive character – a monotonic increase in amplitude to maximum of the seismic moment takes about 5 s (Pirli et al. 2010). In such a case, correlation of further arrivals is practically impossible. This is illustrated in Fig. 9(B) showing synthetic seismograms calculated for an extremely long-source pulse. However, because of the large magnitude of the event (M = 5.9), correlation of the first arrivals was unquestionable and was done manually using the software ZPLOT (Zelt 1994), which allowed flexible use of scaling, zooming, filtering and reduction velocity. For the two northern events at Spitsbergen and Novaya Zemlya, we observe a low-velocity layer, or asthenosphere. The LAB beneath Barents Sea was found at a depth of about 200 km (198 and 201 km for events no. 2 and no. 8, respectively).

The section for event no. 7 (Bełchatów) shows continuous first arrivals of P-waves with no evidence for a “shadow zone” in the whole range of registration (Fig. 9(c)). A lack of “shadow zone” could be interpreted as a lack of asthenosphere beneath central part of the Baltic Shield, or that the LAB occurs deeper in this area (>200 km).

For event no. 7 (Bełchatów), an example of the resolution test of calculated travel times is shown in Fig. 9(c). The arrival times of P-phase were calculated for the lower lithosphere (lid) model perturbed by +0.05 and − 0.05 km s^− 1 relative to the final model (see red curve in Fig. 14(b)). Travel times shown by blue and orange lines are significantly too early and too late. This indicates that the average lid velocity could be determined with an accuracy better than ± 0.05 km s^− 1. Uncertainties in the order of ± 0.05 km s^− 1 are rather small, particularly if we compare it with the results of receiver function or teleseismic tomography. However, in our method, the seismic rays are travelling nearly horizontally in the lid and over long distances in the order of 1000–1500 km. In receiver function and teleseismic tomography studies, rays cross the lid nearly vertically with ray paths of about 200–300 km long only.

Models of the lower lithosphere structure

Our other data do not show evidence for the existence of the asthenosphere beneath the Baltic Shield. However, they contain information about the lower lithosphere heterogeneities, which could be deduced from P- and S-waves (Figs. 10111213).

Fig. 10 Record section (A) for event no. 9 (Rhine Graben) with signals from simultaneous small event in Finland (event no. 10; highlighted by yellow). B. Location of event no. 10 and seismic stations recording this event. C. Record section of event no. 10. Note the good fit of P and S travel times calculated for the reference model.

Fig. 11 Record section (bottom) for event no. 1 (Skagerak) with synthetic seismograms (top). Note in the section a “shadow zone” (at a distance of around 600 km) which is an effect of an LVL in the lower lithosphere (at a depth interval of 61–77 km). Red lines show the first arrivals (Pn and P), reflections from the bottom of the LVL (1) and travel times of five multiples in the LVL (2)–(6). Multiples in the LVL (2)–(4) explain the strong amplitudes in further arrivals in the distance range of 600–800 km.

Fig. 12 A. Differentiation of the upper mantle velocities observed for event no. 4 (Varangerfjord). Low velocities in the northern part of the Baltic Shield (green dots), and normal to the south (red dots). Thick grey bar highlights far-distance Pg waves recorded at distances up to 600 km. B. Location map showing recording stations in corresponding colours.

Fig. 13 A. Differentiation of the upper mantle velocities observed for event no. 5 (Skåne). Normal velocities to the north (red dots), higher to NE (light blue dots) and highest to the east (navy blue dots). Thick grey bar highlights far-distance Pg waves recorded at distances up to 600 km. B. Location map showing recording stations in corresponding colours.

The sections shown in Fig. 10 need more explanation. About 4 min (precisely 257.7 s) after event no. 9 (Rhine Graben, M = 4.0), at a distance of about 1800 km towards the northeast, a smaller earthquake occurred in western Finland (event no. 10, M = 1.1; see Table 1). The section in Fig. 10(a) shows both events, recorded simultaneously. Phases attributed to the small event in Finland are highlighted in yellow (recorded at epicentral distances of 1500–2200 km from event no. 9). In Fig. 10(b)(c), the location of this event and the record section are, respectively, shown. Pg, Pn, Sg and Sn waves calculated for the reference model of the Baltic Shield fit observations quite well. For event no. 9 (Fig. 10(a)), strong P- and S-waves are recorded up to 1500 km distance. However, in the distance range 1600–2300 km, the strong P-waves have no corresponding strong S-waves. This fact could be interpreted as lowering of the S-wave velocity (the S-wave asthenosphere); another explanation could be lower quality factor for S-waves (“Q_s asthenosphere”?).

Lower lithosphere heterogeneities were interpreted beneath southern Sweden using data for event no. 1 (Skagerak). At a distance of c. 600 km, a small break in the continuity of the first arrivals of P-waves is observed (Fig. 11). This can be explained by a model with a 16-km-thick low velocity layer (LVL) 15 km below the Moho and with a P-wave velocity drop of − 0.33 km s^− 1. Such a velocity contrast is sufficient to explain the strong amplitudes, seen in the distance range of 1600–1800 km, as multiple reflections within the LVL. Red lines in Fig. 11 show the first arrivals (Pn and P), reflection from the bottom of the LVL (1) and the travel times of five multiples in the LVL (2)–(6). In synthetic seismograms, reflection from the bottom of the LVL (1) and multiples (2) and (3) are strong, (4) is weak and higher multiples practically invisible. We do not interpret the LVL as the asthenosphere but as an inhomogeneity within the lower lithosphere.

The next record section (Fig. 12(a)) illustrates a differentiation of the upper mantle velocities observed for the event no. 4 (Varangerfjord). Although time corrections have been applied, considerable scattering of the first arrivals is observed, particularly in the distance interval of 200–600 km. Later arrivals are marked in green and corresponding stations are also marked in green in the neighbouring map (Fig. 12(b)). Low velocities in the lower lithosphere coincide with the northern part of the Baltic Shield and “normal velocities” with the area towards the south (red dots). Even larger scattering of the first arrivals, up to about 3 s at about 800 km distance, is observed for event no. 5 (Skåne) shown in Fig. 13. In this case, the first arrival picks were split into three groups: normal arrivals to the north (red dots), faster to the northeast (light blue dots) and the fastest to the east (navy blue dots).

In the last two sections for event no. 4 (Varangerfjord) and for event no. 5 (Skåne), the thick grey bar highlights far-distance Pg waves recorded up to about 700 km distance (Figs. 12 and 13). This means, qualitatively, that attenuation of P-waves in the Baltic Shield crust is relatively low. Beneath the SVEKA profile in Finland (Grad & Luosto 1994), the Q_P-factor in the uppermost 1 km is only 50–80, but in the crystalline crust it reaches values of 500–800, which means that attenuation in the crust is extremely low.

The records of the last two sections (Figs. 12 and 13) reach distances up to about 1700 km, too short to reach the LAB in the central part of the Baltic Shield, where it is at depth deeper than 200 km. On the other hand, both events, located in the northern and in the southern parts of the shield, give valuable information about the lower lithosphere – lid. Based on these data, the Baltic Shield lid can be divided into areas of different P-wave velocities (Fig. 14(a)): normal velocity (approximately corresponding to the area of Sweden), faster velocities (central and southern Finland), the fastest velocities (at the southeastern edge of the shield) and the lowest velocities (coincides with the northern part of the Baltic Shield in northern Finland and northern Norway).

Fig. 14 Summary of the Baltic Shield structure. A. Map showing regions of lithospheric mantle velocities: low in the northern Baltic Shield (green), normal (red), higher (light blue) and highest in the east (navy blue dots). Two red circles in the north of the shield show locations of deepest rays from event no. 2 (Spitsbergen) and event no. 8 (Novaya Zemlya) corresponding to a LAB at 200 km depth in the Barents Sea. LAB in the central Baltic Shield should be deeper than 200–250 km. B. Models of lithospheric mantle velocities for P-waves for event no. 4 NBS(4) – Northern Baltic Shield (Varangerfjord, green line) and event no. 5 (Skåne, red line). Colours correspond to those on adjacent map; iasp91 model (yellow line) is shown for comparison. C. P- and S-wave velocity models for the Baltic Shield together with the iasp91 model (yellow line). Two grey lines for events no. 2 and 8 document LAB and asthenosphere, respectively. The lowest velocities were found for event no. 4 in the northern Baltic Shield. Model for event no. 1 shows a low-velocity zone in the lower lithosphere of the southern Baltic Shield (compare Fig. 11). See text for further discussion.

Summary of the upper mantle structure and comparison with other methods

The seismic structure of the lithosphere–asthenosphere system beneath the Baltic Shield was studied using local earthquakes with magnitudes ranging from 2.7 to 5.9. Good-quality records were obtained up to 2200–3000 km distances by a relatively dense network of seismic stations, with a station spacing of 30–100 km. All the data were corrected for topography and Moho depth for each event and each station location, using a reference model of the Baltic Shield with 46-km-thick crust. A summary of the Baltic Shield structure is compiled in Fig. 14, which shows a map of the velocity provinces (Fig. 14(a)), P-wave velocities in the lid (Fig. 14(b)), and P- and S-wave velocities in the upper mantle (Fig. 14(c)).

In general, P and S arrivals observed in the Baltic Shield are significantly earlier than predicted by the iasp91 global velocity model, which clearly indicates that lithospheric P and S velocities of the Baltic Shield are higher (see Figs. 6(b),(c) and 7(b),(c)). On other cratons, velocities in c. 200 km deep lithosphere consistently reach on average about 5% higher values when compared with global averages (Eaton et al. 2009; Lebedev et al. 2009; Jones et al. 2010).

For the two northern events (Spitsbergen and Novaya Zemlya), we observe a LVL that we interpret as the asthenosphere. The LAB beneath Barents Sea was found at the depth of 198 and 201 km for event no. 2 (Spitsbergen) and for event no. 8 (Novaya Zemlya), respectively. The locations of the deepest points reached by rays in the lithosphere are shown in the map (Fig. 14(a)) by two red circles, and the red-dotted line shows an approximately 200-km isoline of the LAB depth. Other data show that LAB in the central Baltic Shield, if it exists, should be deeper than 200–250 km. These depths coincide with results of surface wave tomography of the Barents Sea and surrounding regions (Levshin et al. 2007), as well as thermal modelling of the EEC (Artemieva 2003, 2007). The asthenosphere thickness for the events at Spitsbergen and Novaya Zemlya was determined only from P-waves, and they are 69 and 61 km, respectively (see grey lines in Fig. 14(c)). Analysis of the other record sections shows that the first arrivals of P-waves are continuous in the whole range of registration, while corresponding S-waves are weaker. Weak S-waves can be explained by lowering the S-wave velocity or decreasing the quality factor for S-waves (S-wave asthenosphere or Q_S asthenosphere, it means a lower S-quality factor).

The thick lithosphere beneath the Baltic Shield was also found from teleseismic data recorded by broadband stations of the Swedish National Seismic Network. Using waves converted from S to P in the uppermost mantle, the LAB was clearly imaged at depths of about 200 km (Olsson et al. 2007). Beneath the Teleseismic Tomography across the Tornquist Zone in Germany–Denmark–Sweden (TOR) array, existence of a major near-vertical lateral change in the P and S velocities of several percents was observed up to depths of 250 km, more or less directly below the edge of the Baltic Shield (Shomali et al. 2006). Shomali et al. (2006) conclude that there is no asthenosphere under the southern Baltic Shield area for which four possibilities can be considered: there may be no conventional asthenospheric layer; the base of the lithosphere is deep and is not resolvable based on the TOR data-set; the transition (LAB) is characterized with a small velocity change and resolution of the base of the model is low. A deep LAB has also been found for other cratons, e.g. beneath the Slave craton in the northwestern Canadian Shield (Chen et al. 2007). The 1D inversion of phase velocities yields high upper mantle S-wave velocities of 4.7 ± 0.2 km s^− 1 that persist to 220 ± 65 km depth and thus define the cratonic lithosphere. The Rayleigh wave phase velocities at the period range of 20–142 s give the greatest sensitivity at depths of 28–200 km. The lithosphere is thinner beneath the eastern margin of the South Australia Craton. Prominent S-phases from a negative velocity contrast were found at depths of 131 ± 9 km, consistent with the LAB depth range derived from surface wave tomography (Ford et al. 2010).

From the analysis of record sections from different areas in the Baltic Shield, the lid of the shield can be divided into areas of different P-wave velocities. In Fig. 14(a), the area of normal velocity is marked in red (approximately corresponding to the area of Sweden), faster velocities in blue (central and southern Finland), fastest velocities in navy blue (at the southeastern edge of the shield) and the area of the lowest velocities is marked in green (coincides with the northern part of the Baltic Shield in northern Finland and northern Norway). The corresponding models of P-wave velocity are shown in the same colours in Fig. 14. Differentiation of the velocity could be interpreted as heterogeneity of the lid. However, another interpretation could be anisotropy of the lithosphere, with the fast velocity direction approximately in the east–west direction and the slow velocity direction approximately in the south–north direction.

The upper mantle structure under the Baltic Shield has also been studied using teleseismic P and S tomography and shear-wave splitting (Eken et al. 2007, 2008, 2010). The results indicate P-wave velocity perturbations of ± 3% down to at least 470 km below the SNSN. Below Central Sweden, where ray coverage is the best, the data reveal a large region of relatively low velocity at depths of c. 300 km. In the TOR experiment, joint analysis of shear-wave splitting parameters and the directional dependence of teleseismic P residuals show different lithosphere thicknesses and different orientations of seismic anisotropy in the mantle lithosphere. The three blocks separated by the major sutures are clearly differentiated by the orientation of large-scale olivine fabric in the mantle lithosphere. Fast velocities in the cratonic lithosphere of the southern Sweden domain are parallel to the craton edge (Babuška & Plomerová 2004; Plomerová & Babuška 2010).

Regional teleseismic tomography carried out within the SVEKALAPKO seismic experiment was designed to study the upper mantle of the Baltic Shield between the Archaean and Proterozoic domains (Bruneton et al. 2004). The 3D shear-wave velocity model obtained from the fundamental mode Rayleigh waves shows lateral variations in velocities down to 150 km depth. The model exhibits variations in the S-wave velocities up to ± 3%; however, they are not noticeably different beneath the Archaean and Proterozoic domains. As the thermal variations beneath this part of shield are very small, these lateral variations must be caused by different rock compositions.

Apart from regional velocity differentiation, we found small-scale lithospheric inhomogeneities. Beneath southern Sweden, 15 km below the Moho, a LVL of 16 km thickness with a − 0.33-km s^− 1 drop of velocity for P-waves was found. Reflected waves and multiples in the LVL agree well with observations (Fig. 11). For the same area of southern Sweden, a lithospheric reflector was found at a similar depth from FENNOLORA data (e.g. Guggisberg 1986), as well as in the uppermost mantle beneath southern Finland from local events registered by the SVEKALAPKO seismic array (Yliniemi et al. 2004). The analyses of data from the TOR experiment all show the Tornquist zone to be a sharp boundary throughout the lithosphere, with a lithosphere thickness of at least 200 km for the southernmost part of the Baltic Shield (Shomali et al. 2006).

Conclusions

• The asthenosphere was found at a depth of c. 200 km in the area north of the Baltic Shield (beneath the Barents Sea).

• No evidence was found for the asthenosphere beneath the Baltic Shield – it has to be deeper than 200 km, if it exists at all. Even if it exists at greater depths, it could not be detected by the “refraction” method. For distances larger than about 2000 km, a “shadow zone” in the first arrivals would be masked by deeper waves from the “410-” and “660”-km boundaries.

• Difficulties in the interpretation of later arrivals result from the complicated source function of natural earthquakes (e.g. event no. 2) and from multireflections in the LVL (event no. 1). Interpretation can also be complicated by seismic noise when signals from distant earthquakes are not strong enough or obscured by simultaneous recording of small local events (see events no. 9 and 10).

• Differentiation of the velocity in the lid could be interpreted as regional heterogeneity or anisotropy; however, more 3D tomographic studies are required.

Acknowledgements

The authors thank Finnish Academy of Science and Letters, Väisälä Foundation, for financial support. This work was partially supported by NCN grants UMO-2011/01/B/ST10/06653 and DEC-2011/02/A/ST10/00284. The authors are also grateful to the staff at the SNSN for giving the authors access to their data. Waveform data from seismic stations operated by the Institute of Seismology of the University of Helsinki, the Sodankylä Geophysical Observatory of the University of Oulu and from the SNSN were used in this study. For specific events, data from Norwegian and Russian stations were made available by the Bergen Seismological Observatory and NORSAR in Norway. The authors wish to thank GEOFON, GFZ German Research Centre for Geosciences for earthquake information and waveform data. The authors also thank two referees, Prof. C. Juhlin and Prof. R.W. England, for their helpful comments. The authors are grateful to Dr Emilia Koivisto and editor Dr Magnus Ripa for improving the English language. Geographic data handling and plotting were done with the GMT software by P. Wessel and W.H.F Smith (Wessel & Smith 1991, 1998).

Notes

^* Current affiliation: Department of Mineral Resources, Geological Survey of Sweden, P.O. Box 670, SE-75128 Uppsala, Sweden