Lithosphere Structure in Europe from Thermal Isostasy T Irina M

Earth-Science Reviews 188 (2019) 454–468

Contents lists available at ScienceDirect

Earth-Science Reviews

journal homepage: www.elsevier.com/locate/earscirev

Lithosphere structure in Europe from thermal isostasy T Irina M. Artemieva

IGN, University of Copenhagen, Denmark

ARTICLE INFO ABSTRACT

Keywords: A new method for modeling the thermal structure of the lithosphere, termed thermal isostasy, is presented and Lithosphere thermal thickness used to calculate the lithosphere thickness and to predict geothermal heat flux in Europe and the adjacent off- East European craton shore regions. The method is based on analysis of topography deviations from the expected correlation between Barents Sea the Moho depth and topography. Anomalous topography is interpreted as caused by thermal anomalies in the Variscan Europe lithosphere and by variations in the lithosphere thermal thickness. The new method allows for recovering the Collisional orogens lithosphere thermal structure with a high lateral resolution and for testing the results by comparing the predicted North Atlantic Ocean and measured heat flux. The results for Europe, constrained by the regional seismic crustal model EUNAseis, demonstrate that the lithosphere thermal structure is controlled by both the geodynamic setting and the lithosphere age. The lithosphere of the East European craton ranges in thickness from ca. 140-180 km in the paleorifts to 200-220 km on average with local lithosphere roots down to ca. 300 km depth, such as in the south-central Proterozoic Finland, and local thinning to < 100 km in the Peri-Caspian Basin and the Mezen rift system. Lithosphere thinning to 70- 100 km beneath southern Norway may have caused its recent uplift. The craton to noncraton transition at depth roughly follows the geological boundary but with up to 200 km lateral deviations, which indicate reworking of the cratonic lithosphere along the craton edge. Paleoproterozoic – early Paleozoic Gondwana massifs of western Europe are distinct in having cold and thick lithosphere, typically 120-180 km thick, up to 200 km thick in the London-Brabant Massif, and only ca. 100 km in the Iberian Massif. The European Cenozoic Rift System is marked by a thin (< 80 km) and hot lithosphere that forms a linear belt from southern France to the North Sea. Lithosphere thinning to 50-60 km thickness beneath the Pannonian and Po basins may indicate the onset of oceanization. Cenozoic orogens of western Europe have lithospheric roots down to ca. 200-250 km associated with subducting lithospheric slabs, while a 80 km thick lithosphere beneath the Vrancea zone is in agreement with slab delamination. The Arctic shelf of the Barents Sea has a thick cratonic lithosphere with a sharp transition from a 120-150 km thick lithosphere of the western Barents Sea to a 175-230 km thick lithosphere of the eastern Barents Sea. The block with a ca. 150 km thick lithosphere in the North Atlantic region between the Aegir paleo-spreading and Norway may represent a continental terrane within the oceanic domain. The upper mantle temperature beneath the Bay of Biscay and the Black Sea is similar to the cratonic and is anomalously cold down to a 200 km depth.

1. Introduction components: structural heterogeneity of the crust and of the lithospheric mantle, and both layers include variations in thickness and Isostasy is one of the fundamental concepts in geophysics. Most of density. Therefore, in regions where the crustal structure is known from the Earth is close to isostatic equilibrium, and the largest deviations seismic observations, isostasy provides constraints on the structure of from regional isostasy with strong free air anomalies are associated the lithospheric mantle. with the subduction systems (Watts, 2001). The compensation depth Crustal thickness is constrained seismically for many regions and where isostatic equilibrium is achieved is not well-known, and the crustal density is known from laboratory measurements on rock sam- classical concepts of Pratt and Airy assume that isostasy is achieved at ples from different crustal layers, which also provide relationships be- the base of the crust. However, it is likely to be deeper and it is usually tween seismic velocities and rock densities (Ludwig et al., 1970; expected that the compensation depth is not deeper than the base of the Carlson and Herrick, 1990; Christensen and Mooney, 1995). Density lithosphere. Thus isostasy is maintained by an interplay of several heterogeneity of the lithospheric mantle is caused by variations in both

E-mail address: [email protected]. https://doi.org/10.1016/j.earscirev.2018.11.004 Received 19 August 2018; Received in revised form 25 October 2018; Accepted 6 November 2018 Available online 16 November 2018 0012-8252/ © 2018 Elsevier B.V. All rights reserved. I.M. Artemieva Earth-Science Reviews 188 (2019) 454–468 temperature and composition, and except for the cratonic regions with deep (Shapiro and Ritzwoller, 2002; Zhu et al., 2012; Schaeffer and a uniquely depleted composition (Jordan, 1988; Boyd, 1989; Forte and Lebedev, 2013). Yet, seismic anisotropy provides a mean value of Perry, 2000; Carlson et al., 2005; Lee, 2006; Griffin and O'Reilly, 2007; 170–180 km for the cratonic LAB in Europe (Plomerova and Babuska, Artemieva and Vinnik, 2016) and relatively small variations in thermal 2010), in contrast to a ca. 200 km thick lithosphere based on RFs structure (Artemieva, 2006), temperature variations play a critical role (Vinnik et al., 2016), and a ca. 250 km thick electrical lithosphere on in lateral variations in lithosphere mantle density (Deschamps et al., average, with local LAB depth anomalies down to ca. 400 km depth 2002; Simmons et al., 2009). Thus, one may calculate the lithosphere (Korja, 2007). It is, however, still debated if seismic RFs image the LAB thermal structure (and therefore the lithosphere thermal thickness), (Rychert and Shearer, 2009) or instead an intra-lithosphere boundary assuming that compositional density variations in the lithospheric (e.g. MLD, the mid-lithosphere discontinuity) with a high velocity mantle play minor role in lithosphere isostasy. As discussed below, this contrast (Romanowicz, 2009; Karato et al., 2015). There is also a sig- is also often true for the cratonic lithosphere. nificant inconsistency between regional seismic tomography models for Here I describe a new method for calculating thermal structure of all parts of Europe (Yang et al., 2007; Zhu et al., 2012; Schaeffer and the lithosphere, which I apply to Europe (Fig. 1). In a companion paper, Lebedev, 2013; Lebedev et al., 2018). the method has been applied to Greenland where it demonstrates the Thermal structure of the European lithosphere has been addressed efficiency in calculating lithosphere thermal structure and surface heat only in few studies. The cratonic part of Europe is covered by a global flux even in regions with sparse geophysical data(Artemieva, 2018). model, where thermal structure of the lithosphere is calculated from The European region is covered by a high-resolution detailed surface heat flux data (Artemieva and Mooney, 2001). Since heat flux seismic database, EUNAseis, on the crustal structure (Artemieva and measurements are absent for large parts of the continents, for regions Thybo, 2013) which allows for an accurate calculation of the crustal with no borehole measurements the global thermal model for the Pre- contribution to the isostasy. Most of the region has near-zero free air cambrian lithosphere (Artemieva and Mooney, 2001) has been com- gravity anomalies (Fig. 2a), with significant deviations ( ± 100 mGal) plemented by statistical data and for active regions thermal constraints only in the Alps, the Caucasus, and the Mediterranean region. There- from geothermobarometry were used. This way the global continental fore, the proposed approach is valid for most of the European continent. model TC1 was constrained (Artemieva, 2006). These are the only ex- The European region is also densely covered by geophysical ob- isting large-scale models for thermal structure of the European litho- servations with a large number of seismic, gravity, thermal, and elec- sphere, except for few 30–40 years old results for the European con- tromagnetic regional models of the lithosphere structure (reviewed in tinent (Cermak and Rybach, 1979) and a small number of regional-scale Artemieva et al., 2006), which do not always agree. Significant differ- studies (e.g. Cermak and Bodri, 1995; Norden et al., 2008). However, ences between the lithosphere thickness estimates in Europe based on shallow upper mantle temperatures, mostly under the Phanerozoic part seismic and magneto-telluric studies motivated a group of researchers of Europe, have also been constrained by seismic tomography with to introduce several new definitions of the lithosphere-asthenosphere significantly different results for mantle temperatures based ontheP- boundary (LAB), depending on the geophysical method employed and S-wave models (Goes et al., 2000). (Jones et al., 2010). For example, in Phanerozoic Europe seismic re- This situation naturally calls for new constraints on the lithosphere ceiver functions (RFs) and magnetotellurics are consistent in de- thermal structure in Europe. Here I propose the new thermal isostasy termining the LAB depth of 90–100 km (Praus et al., 1990; Geissler method which also allows for predicting surface heat flux. A dense et al., 2010), while seismic anisotropy indicates it is around 135 km network of borehole heat flux measurements in Europe provides an thick (Babuska et al., 1998; Plomerova et al., 2002), and most seismic opportunity for comparing model predictions with observations and tomography models indicate that it is substantially less than 70–100 km therefore for direct testing of the new method.

Fig. 1. Tectonic map of Europe (modified after Artemieva and Thybo, 2013).

455 I.M. Artemieva Earth-Science Reviews 188 (2019) 454–468

Fig. 2. Gravity anomalies based on the EGM2008 global gravity model (Pavlis et al., 2012). (a) Free air anomalies, (b) Bouguer anomalies. Thin lines – boundaries of major tectonic provinces (simplified after Artemieva and Thybo, 2013).

2. Thermal isostasy method regionally in specific geodynamic settings (Cloetingh et al., 2007). • Compositional heterogeneity of the lithosphere mantle is important 2.1. Assumptions for the cratonic regions, where it is well established from petrolo- gical data (Gaul et al., 2000; Poudjom Djomani et al., 2001; Carlson The proposed method is based on the following assumptions: et al., 2005) and geophysical studies (Artemieva, 2009; (1) The study region is close to isostatic equilibrium (Fig. 2a). For the Cherepanova and Artemieva, 2015; Artemieva and Vinnik, 2016). continental lithosphere, for simplicity I first assume Airy-type isostasy Depletion of cratonic lithosphere mantle in basaltic components with the compensation level at the Moho. It implies that there should be results in density decrease (Lee, 2003; Carlson et al., 2005). At room a linear correlation between the Moho depth and the topography (e.g. P-T conditions it may amount to ca. 0.05–0.07 t/m3 in shallow li- Fowler, 2005). The slope of the correlation between the two parameters thosphere mantle (Jordan, 1981; Poudjom Djomani et al., 2001). depends on the density contrast across the Moho, while the intercept Since depletion in cratonic lithosphere mantle usually decreases defines average crustal thickness at zero topography. By comparing re- with depth due to metasomatic enrichment, a maximum difference gional data for the Moho depth and topography with the best-fit corre- in bulk lithosphere mantle density between a highly depleted Ar- lation between the two parameters, one can test the assumption that chean and a fertile Phanerozoic lithosphere mantle probably may isostasy is indeed achieved at the Moho. If this is the case, the data will not exceed 0.03–0.05 t/m3. This variation has little effect on the plot around the theoretical best-fit line. However, analysis of real data for calculated temperature anomalies (Fig. 4 and discussion below). any region in Europe shows that only a small part of the data follows the Therefore, I adopt the assumption that temperature variations in the theoretical prediction. Therefore, the assumption on isostatic compen- lithosphere have the primary effect on mantle density anomalies sation at the Moho is not satisfied and the deviations of data from pre- (Deschamps et al., 2002; Schutt and Lesher, 2006; Simmons et al., dictions require a different style of compensation. I next assume that 2009). I demonstrate that these simplifications are justified within isostasy is fully satisfied at the lithosphere base (the LAB). the model uncertainty and provide reliable results that agree with (2) I interpret local deviations from the crustal isostasy predictions independent observations, such as borehole heat flux measurements. in terms of “anomalous topography”. I assume that the effect of crustal Further discussion can be found in Artemieva (2018). density heterogeneity is compensated by the Moho depth variations. Regional analysis for the European continent with a highly hetero- (4) I therefore assume that isostasy is fully satisfied at the LAB, that geneous crustal structure (Fig. 3) demonstrates that this assumption anomalous topography is caused by density anomalies in the litho- leads to realistic results for the lithosphere thermal structure. sphere mantle and by variations in lithosphere mantle thickness, and (3) Principal reasons for topography deviations from the crustal that density anomalies in the lithospheric mantle are caused by tem- isostasy predictions are lithosphere mantle density heterogeneity of perature variations alone. This assumption allows for calculating the thermal and/or compositional origin and lithosphere deformation. lithosphere thermal thickness and provides a means of predicting geothermal heat flux. An additional assumption is required for the litho- • Lithosphere deformation depends itself on the thermal state of the sphere thickness value at locations where anomalous topography lithosphere (Forte and Mitrovica, 1996; Cloetingh et al., 2004), A = 0 km (explained in the next section). Such locations are referred to while the goal of this study is to determine the latter. I therefore below as “reference points”. ignore possible effects of flexural isostasy (Beaumont, 1981; Watts, (5) Geothermal heat flux is calculated based on continental re- 2001) and horizontal forces caused by differences in gravitational ference geotherms (Pollack et al., 1993). The results are as good as potential energy between adjacent lithospheric blocks (Artyushkov, model assumptions used to constrain reference geotherms, that is 40% 1973; Fleitout and Froidevaux, 1982). Within the European con- of surface heat flux is generated by radiogenic decay in the upper 8km tinent, the contributions of these factors may be important only of continental crust (Pollack and Chapman, 1977).

456 I.M. Artemieva Earth-Science Reviews 188 (2019) 454–468

Fig. 3. Crustal structure based on regional seismic model for the crust EUNAseis (Artemieva and Thybo, 2013). (a) Moho depth, (b) Average density in the crust (including sedimentary strata). For seismic data coverage see Fig. 6b. For regions without seismic data, crustal thickness is constrained by interpolation for continental crust (including shelves) and a standard value of 7 km is assigned to oceanic crust. Seismic profiles in the south-eastern part of the study region (the Caspian region, the Black Sea, and the Turanian plate) are mostly from the 1970-ies and their quality is unknown.

anomaly ΔT is averaged over the entire lithosphere column and it is defined with respect to an average temperature profile at the reference points where A = 0 km. Since

T= o (1–T ), (2) where α = 3.5 ∙ 10−5 1/K is the coefficient of thermal expansion, combination of Eqs. (1) and (2) yields

TA= m/.zLAB o (3) Therefore, by using Eq. (3)A can be directly converted to thermal anomalies in the lithosphere ΔT, assuming that all other contributions

to the anomalous topography A are negligible and zLAB is known. The latter requirement is critical, and as demonstrated below it can be determined in a self-consistent way from A. The following parameters are required to calculate ΔT:

i. The A-values are calculated from a correlation between the Moho depth and topography. The calculated best-fit regional correlation between these two parameters reflects regional characteristics of (a) Fig. 4. Effect of mantle lithosphere density variations on calculated tempera- the crustal density structure through the slope of the best-fit line ture anomaly ΔT (eq. 6). A 1 km anomalous topography can be explained by a and (b) typical regional patterns in the crustal thickness through the 150–200 °C temperature anomaly in lithosphere mantle. The result is nearly intercept of the best-fit line. As the example for Europe shows, the independent from density variations in the mantle lithosphere and the crust, data may require splitting regions into subdomains. In particular, but depends on lithosphere thickness. the crust of western Europe is thin (Fig. 3a) and has low density (Fig. 3b) because it mostly lacks the lower crustal layer, in contrast 2.2. Temperature anomalies in the lithosphere to the thick three-layer crust on the East European craton (Artemieva and Thybo, 2013). Assumptions 1 and 2 allow for calculating deviation (A) of true ii. Asthenosphere density at T = 0 °C (standard pressure-temperature 3 topography from the Airy isostatic prediction at each location, based on laboratory conditions, SPT) is assumed to be ρm(0) = 3.40 t/m , the correlation between the Moho depth and the observed topography which corresponds to fertile mantle composition (Gaul et al., 2000). 3 in the study region. According to assumption 4, anomalous topography This value yields in situ ρm=3.28 t/m assuming a Moho tempera- A is caused by variations in lithosphere temperature: ture of 600 °C and an asthenosphere temperature of 1400 °C. iii. Lithosphere (crust plus lithosphere mantle) density at T = 0 °C is A= zLAB [–]/o T m (1) o (0)= [2.85zMoho + 3.35(zLAB – z Moho )]/ z LAB (4) where zLAB is the depth to the lithosphere base (lithosphere thermal where zMoho and zLAB are the depths to the Moho and to the LAB, re- thickness), ρo is in situ density of the lithosphere (crust plus mantle) spectively, with the average SPT densities of 2.85 t/m3 for the bulk when A = 0 km, ρm is in situ density of the sublithospheric mantle crust and 3.35 t/m3 for the bulk lithosphere mantle (the latter corre- (asthenosphere), and ρΔT is in situ lithosphere density at temperature sponds to a typical density of Proterozoic lithosphere mantle, Gaul anomaly ΔT (i.e. it is this anomaly that produces A). Temperature

457 I.M. Artemieva Earth-Science Reviews 188 (2019) 454–468 et al., 2000)). Here I use the density of Proterozoic mantle as a compromise intermediate value, given that the volume of lithospheric mantle of different ages and the extent of its chemical modification by magmatism are not well known (Artemieva, 2009), and the European continent includes both depleted cratonic lithosphere and young fertile lithosphere. For the same reason, I use a compromise intermediate value for the bulk crustal density of 2.85 t/m3. Its choice has negligible effect on the results (see Appendix A for the uncertainty analysis), while the choice of the SPT lithosphere mantle density slightly shifts the final values (Fig. 4). Assuming surface temperature of 0 °C and temperature at the li- Fig. 5. Sketch illustrating temperature anomaly ΔT in the lithosphere (averaged thosphere base of 1400 °C, the average lithosphere temperature is from the surface down to the LAB) with respect to a reference point, where 700 °C. Then in situ lithosphere density is: A = 0 km and the geotherm follows black dashed line. At locations where the o= o (0)(1–700 ) (5) lithosphere is thinner than at reference point, the geotherm down to the LAB depth at reference point includes the lithospheric part (red line) and the asth- For example, in case crustal thickness is 40 km and lithosphere enospheric part (magenta line). At locations with a thick lithosphere, the geo- 3 3 thickness is 200 km, ρo(0)=3.25 t/m and ρo=3.17 t/m . therm (blue line) is colder than at reference point, and therefore average li- Substituting all parameters in Eq. (3) by the values defined by Eqs. thosphere temperature is colder than at reference point. (4) and (5) yields:

T= 94683 A /(3.35 zLAB – 0.50 z Moho ) (6) and TT=–700 = 700(1–z /200) Here the A-value and thicknesses of both the lithosphere and the ave LAB (8) o Ref crust are in km. Numerical values in Eq. (6) depend on the assumed where ΔT is in C and 200 (in km) is zLAB at reference points where crustal and mantle densities (Eqs. 3–5) and would vary with different A = 0 km. A combination of Eqs. (6) and (8) allows for calculating li- assumptions. For example, for the average SPT densities of the bulk thosphere thickness zLAB from the anomalous topography A: crust and the bulk lithosphere mantle of 2.80 t/m3 and 3.40 t/m3, re- A= (200 zLAB ) (3.35 zLAB 0.50zMoho )/27029. (9) spectively, Eq. (6) will take the form: ΔT = 96098 A/(3.40 zLAB – 0.60 zMoho). Eq. (6) implies that the anomalous equivalent topography of Importantly, the relationship (9) between A and zLAB is non-linear A = 1 km requires a lithosphere temperature anomaly of ca. 145 °C (see Artemieva, 2018 for details). The physical reason is the competing with respect to reference points in case of a 200 km thick lithosphere effects of variations in zLAB values on average density of the crust- with a 40 km thick crust (Fig. 4). Fig. 4 based on Eq. (6) also illustrates mantle column above the reference LAB depth (here 200 km): a thin that the choice of the mantle lithosphere density has little effect on the lithosphere implies high average temperatures and the associated calculated temperature anomaly. Further discussion of model sensi- density decrease due to rock thermal expansion; this is compensated by tivity and resolution is in Appendix A. a simultaneous increase in the thickness of a denser (than the lithosphere column) sublithospheric mantle layer between the local LAB and iv. The Moho depth is usually known from seismic data, and variations the compensation depth (the LAB at reference points). Importantly, due

in this parameter, especially in case of a thick lithosphere, have little to the non-linear relationship between A and zLAB in Eq. (9), lithosphere effect on the calculated ΔT. Lithosphere thickness variations, how- thickness zLAB cannot be determined when the anomalous equivalent ever, have a crucial effect on ΔT but they are not well known be- topography A is extremely high (> 0.8 km, > 1.2 km, and > 2.0 km Ref cause the existing regional seismic tomography models remain for zLAB =150 km, 200 km, and 250 km, correspondingly). In these highly inconsistent (Schaeffer and Lebedev, 2015). Fortunately, the regions mechanism(s) other than purely thermal contribute to a high lithosphere thermal thickness is directly linked to the thermal topography and large A.

structure of the lithosphere, and zLAB can be determined in a self- The zLAB values determined from Eq. (9) can then be used in Eq. (6) consistent way from the anomalous topography. to calculate temperature anomalies ΔT in the lithosphere. Therefore, the calculation of regional temperature anomalies and regional varia- 2.3. Lithosphere thermal thickness, geotherms and geothermal heat flux tions in the lithosphere thickness requires only independent information on the lithosphere thickness at reference points. Clearly, the choice Ref The temperature anomaly ΔT defined by Eqs. (3) and (6) is a local of another zLAB value will require small adjustments to Eqs. (7)–(9) deviation of vertically averaged lithosphere temperature (from the without changing the general pattern. surface down to the LAB) from averaged temperature at reference Reference geotherms in different tectonic settings (both oceans and points where A = 0 km. It means that at each location the local tem- continents) can be specified by geothermal heat flux(Pollack et al., perature profile is compared to the reference temperature profile. Letus 1993), while in turn reference geotherms define the thickness of the Ref assume for simplicity that at reference points zLAB = 200 km. At lo- thermal lithosphere. Therefore, geothermal heat flux can be estimated cations where the lithosphere is thinner than 200 km, the temperature from zLAB determined at the previous steps. Then lithosphere tem- profile down to 200 km depth includes two parts(Fig. 5): the litho- peratures can be calculated either as depth slices or as vertical profiles sphere part with average temperature of 700 °C (half-temperature be- (geotherms). Note that continental geotherms depend on assumptions tween the surface and the LAB) and the asthenospheric part at tem- on the radiogenic heat production in the crust, and reference con- perature of 1400 °C at depths between the local LAB and the reference tinental geotherms (Pollack and Chapman, 1977; Pollack et al., 1993) LAB, assumed to be at 200 km depth. Likewise at locations with a thick are based on simple assumptions on the ratio of the crustal to mantle lithosphere (> 200 km), the depth of 200 km is within the lithosphere, contribution to surface heat flux, as well as on assumptions on a steady- and therefore temperature at this depth is < 1400 °C, and average li- state thermal regime and purely conductive heat transfer, which may thosphere temperature in the layer above 200 km depth is < 700 °C. both be invalid in active continental regions. Therefore, for an arbitrary location, average temperature in the upper 200 km is 3. Application to Europe: anomalous topography

Tave =[700 zLAB + 1400 (200 –zLAB )]/200 (7) Here the new method of thermal isostasy is applied to the European

458 I.M. Artemieva Earth-Science Reviews 188 (2019) 454–468

Fig. 6. (a) Anomalous topography A calculated as deviation of real topography (b) from topography predicted by isostasy, based on the Moho depth (Fig. 3a) versus topography relationships. Three regional correlations are used in the calculation. Seismic data coverage used in the EUNAseis crustal model (Artemieva and Thybo, 2013) is shown in (b). continent (5 oW – 65°E, 35 – 78 oN) and the adjacent oceanic domains. low-density crust with average density of ca. 2.70–2.75 t/m3 (Fig. A1), such Data on topography is derived from the global ETOPO1 model and is as predicted from the Vp velocity structure of the Variscan crust (Artemieva resampled on a 30’x30’ grid to bring it to the resolution of the crustal and Thybo, 2013)(Fig. 3b). model. All further analysis is presented on a 30’x30’ grid. The East European craton nearly lacks topography, while it has huge The Moho depth (Fig. 3a) is constrained by the EUNAseis regional variations in crustal thickness (Fig. 3a); therefore there is no correlation seismic crustal model (Artemieva and Thybo, 2013), which is based on between the Moho depth and topography in the cratonic lithosphere. For all available regional seismic reflection/refraction profiles digitized this tectonic domain, a relationship zMoho = 38.0 + 5.1 Topo (with all with a minimum of 50 km spacing and complemented by seismic re- parameters in km) is adopted, based on global averages for the continental ceiver functions data on the Moho depth in regions with poor seismic crust (Christensen and Mooney, 1995). The slope is based on the assumption data coverage (Fig. 6b). The cratonic crust of the East European craton of average density of 2.80 t/m3 for the crust and 3.35 t/m3 for the mantle is typically 40–50 km thick, with a thick crustal root (ca. 60 km) in (Fig. A1). The intersect between the best fit line and the topography axis central Finland and some parts of the East European craton. The crust provides a global average depth to the continental Moho (39.2 km) for thins along the paleo-rifts (35–42 km) and in the Barents Sea (ca. average continental topography of ca. 0.24 km. 30–35 km), which forms a part of the Arctic continental shelf. In most of For the ocean domain, the correlation between the Moho depth and western Europe, except for the orogens (Alps, Carpathians, Dinarides, topography is poor (R2 = 0.22) and the data show a large scatter. To Pyrenees), the crustal thickness is ca. 28–32 km and a lower crustal improve model constraints, the domain should preferably be split into layer is nearly absent as reflected in the average crustal velocities smaller regions, however the sparse data coverage in significant parts of (Fig. 3b). Although oceanic lithosphere (defined here by bathy- the ocean domain (the Mediterranean and the Black seas) does not metry < −0.5 km) is also included in the analysis, significant parts of allow for doing this. The best fit relationship is zMoho = 21.6 + 3.5 Topo the oceanic domain, especially in the Mediterranean and the Black Sea, (with all parameters in km), that is a 3 km deep ocean basin has a ty- lack reliable seismic data on the Moho depth, except for some near- pical Moho depth of ca. 11 km. The slope of the best-fit line of 3.5 coast locations. Similarly, seismic data on the Moho depth is nearly requires a small density contrast across the Moho (Fig. A1), due to ei- absent in the south-eastern part of the continental domain, including ther a dense crust, or a low density mantle, or a combination of both. the South Caspian Sea region. For this reason, the modeling results for The anomalous topography A (the deviation of real topography these regions are poorly constrained. from topography predicted from the Moho depth based on isostasy) is For the entire European continent, the Moho depth and topography calculated separately for the same three domains as described above. Its are not correlated (the correlation coefficient2 R = 0.07). Therefore, pattern clearly reflects major tectonic structures in Europe (Fig. 6a). three major tectonic domains are analyzed separately: The cratonic areas generally have negative anomalous topography with typical values of ca. -1 km, except for a few areas with A~1 km • Phanerozoic western Europe plus the Anatolia, the Southern Caspian (southern Norway, the Peri-Caspian depression and the rifts at the region, and the shelf (defined here by bathymetry less than 0.5 km), White Sea coast). Local anomalies of −3 km in central Finland, parts of • the East European craton, including the Baltic Shield, the Volga-Uralia and the Ukrainian Shield are associated with an • the ocean domain, including the Mediterranean and the Black Sea. anomalously thick crust. The Moesia Platform at the SW margin of the craton clearly appears as a separate block with A~0 km. For Western Europe there is a general correlation between the Moho The transition from the cratonic to Phanerozoic Europe is marked by depth and topography, and data for the shelf match the same trend. The best- a sharp change from negative to positive A-values along the Trans- fit line for Phanerozoic Europe is zMoho = 33.4 + 4.3 Topo, where the Moho European Suture Zone (TESZ) and the Svecofennian boundary in the depth and the topography (Topo) are in km. The slope of 4.3 of the best-fit Baltic Shield. This suggests that at lithosphere depths, the edge of the line between zMoho and topography requires a normal mantle density and a craton mostly follows its surface manifestation established from

459 I.M. Artemieva Earth-Science Reviews 188 (2019) 454–468 geological data (Fig. 6a), and that the transition from negative to po- (Fig. 7) and seismic tomography models (Fig. 8), although in most of sitive A-values may be interpreted as the edge of the cratonic litho- the East European craton tomography indicates a deeper LAB than the sphere. In this case, the lithosphere of southern Norway, most of the thermal isostasy model. A discussion of the reasons for the dis- Norwegian Caledonides, the rift system around the White Sea (the area crepancies between different types of geophysical models can be found of the Arkhangelsk kimberlite province), and the anomalous Peri-Cas- in Artemieva (2009, 2011). pian depression (possibly formed as result of the Devonian rifting) has An important difference between Fig. 7a, b is in the lateral dimen- lost its cratonic signature. At the same time, the lithosphere of the East sions of the resolved anomalies. While the thermal isostasy model Barents Sea appears as cratonic. preserves resolution of crustal anomalies with steep gradients, the Most of Phanerozoic Europe has weakly positive A-values (around thermal model recovers a smooth variation in the lithosphere thickness. +0.5 + 1 km) with strong anomalies (A > +1.5 km) in regions of In case the lithosphere deforms as a large-size plate, flexural deforma- active Cenozoic deformation and magmatism (the Rhine Graben, the tion would be important and anomalies should be smoother than cal- French Massif Central, the Pannonian Basin, and the Central Graben of culated from thermal isostasy (Fig. 7a). Yet lithosphere-scale faults, the North Sea). The highly seismogenic Vrancea zone at the bend of the such as expected at boundaries between tectonic terranes and at plate Carpathians, where lithosphere delamination is expected from the boundaries (the craton edge, the Cenozoic orogens formed by colli- seismicity pattern and numerical modeling (Wortel and Spakman, sional processes) will produce zones of sharp gradients in the litho- 2000; Lorinczi and Houseman, 2009; Ren et al., 2012), appears as a sphere thickness as observed in Fig. 7a. separate block with A~ +1.8 km. In contrast, the early Paleozoic Ar- The wavelength of the LAB lateral variations is different in cratonic morican and Wales-London-Brabant Massifs have negative A~ and non-cratonic Europe, with a typical range of 200–300 km within −0.8 km, similar to the lithosphere of the East European craton. All the craton and ca. 100 km in the Phanerozoic part. The boundaries Cenozoic orogens (the Alps, Pyrenees, Dinarides, Carpathians, Cau- between distinct lithospheric blocks are often sharp, such as step-wise casus) have a strongly negative (−2–3 km) anomalous topography; changes in the LAB depth of ca. 150 km over a distance of ca. 200 km at however, the amplitude is highly variable. the edges of the Cenozoic orogens. The western margin of the craton is In the ocean domain, the strongest positive anomaly (A > +3 km) marked by a change in the LAB depth, yet the expression of the craton- is associated with the North Atlantic spreading. Two regions have noncraton boundary in the LAB variation is not very distinct and it strong negative A-values (< −3 km): the Bay of Biscay and the Black follows its geological expression (the Trans-European Suture Zone) only Sea, while the Mediterranean Sea has a patchy pattern of the anomalies. in general, with significant deviations (by ca. 200 km) in southern However, the details are speculative due to poor knowledge of the Norway, Poland and Ukraine (Fig. 7a). These deviations suggest a sig- crustal structure in these regions. nificant reworking of the cratonic margin at depth. In the southern Baltic Shield, the edge of the cratonic mantle approximately follows the 4. Lithosphere thickness in Europe western margin of the Trans-Scandinavian Igneous belt.

4.1. General comments 4.2. The craton

The calculation of the lithosphere thickness requires an assumption In general, the lithosphere thickness in the East European craton is on the LAB depth at reference points with zero anomalous topography. greater than 150 km and the estimated values are in a close agreement In the cratonic part, the zone of near-zero anomalous topography fol- with the lithosphere thermal thickness constrained by surface heat flux lows the western margin of the Svecofennian boundary in the Baltic (Artemieva, 2006). The deepest lithosphere roots (250–300 km) in Shield, where the lithosphere thickness is relatively well constrained by central Finland correlate with constraints from thermobarometry based seismic data (e.g. Pedersen et al., 2013; Silvennoinen et al., 2016; on mantle-derived xenoliths from several kimberlite fields in the region Hejrani et al., 2017). In western Europe, the near-zero anomalous to- which yield a lithosphere thickness of 220–250 km (Kukkonen et al., pography is at the margin of the Armorican Massif, the lithosphere of 2003; Lehtonen and O'Brien, 2009)(Fig. 9). The presence of a more which has been mapped in detail by regional seismic tomography than 200 km deep lithosphere root in southern Finland has also been (Judenherc et al., 2002; Soomro et al., 2016). Based on seismic models, detected by regional seismic tomography (Pedersen et al., 2013) and a value of 150 km is adopted for the LAB depth at reference locations seismic receiver functions (Vinnik et al., 2016). both in the cratonic and non-cratonic Europe, and a value of 100 km is Regional seismic tomography images a patchy pattern of seismic ve- used for reference locations in the ocean domain. locity anomalies, and possibly of the LAB depth, in northern Finland The results (Fig. 7a) show a highly variable thickness of the con- (Silvennoinen et al., 2016). Yet the present study indicates a flat LAB at ca. tinental lithosphere in Europe that ranges from < 70 km in regions of 200 km depth, in agreement with large-scale tomography models (Shapiro Cenozoic magmatism and strong lithosphere extension to 250–300 km and Ritzwoller, 2002; Lebedev et al., 2018)(Fig. 8). It suggests that in parts of the East European craton. There is an overall agreement seismic velocity heterogeneity of lithosphere mantle of the northern Baltic between the lithosphere thickness estimates based on thermal isostasy Shield may be caused by mechanisms other than thermal. In particular, and calculated from surface heat flux measurements (Artemieva and mantle metasomatism at depths from 70 km to ca. 150 km can explain Mooney, 2001; Artemieva, 2006)(Fig. 7b). Most differences are in the average Rayleigh-wave dispersion curves for different cratons, including south-eastern part of the region (the Caucasus region and the Turanian the Baltic Shield (Eken et al., 2018). This interpretation is consistent with plate) where both models are poorly constrained due to scarce data an earlier proposed interpretation that metasomatism of lithospheric coverage: the Moho depth is insufficiently well known for thermal mantle has a strong effect on seismic velocities in the Baltic Shield isostasy modeling, and scarce heal flux data and non-steady state (Bruneton et al., 2004) and cratonic mantle globally (Artemieva, 2009). thermal regime put limitations on thermal modeling. The central and northern parts of the Baltic Shield upper mantle In western Europe, the disagreement between the TC1 thermal show strong seismic anisotropy with significant variations in amplitude model and the present thermal isostasy model is ca. ± 20 km, which is and the fast axis direction between different lithosphere domains within the uncertainty range of the models. Shelf regions cannot be (Plomerova et al., 2011). The anisotropy pattern also changes with compared because they are not covered by the TC1 model. In the depth, and three lithospheric layers are recognized with the boundaries oceanic part of the region, significant discrepancies between the present at ca. 110 km and 220 km (Vinnik et al., 2014). The present results results and the HSC model are around volcanic islands and in regions suggest that the lower layer is located below the base of the thermal with anomalous bathymetry, that does not follow the HSC (half-space lithosphere and its anisotropy may be associated with deformation by cooling) model. There is also an overall agreement between the thermal mantle flow.

460 I.M. Artemieva Earth-Science Reviews 188 (2019) 454–468

Fig. 7. Lithosphere thermal thickness in Europe. (a) Present study, calculated (eq. 9) from the anomalous topography (Fig. 6) and the Moho depth (Fig. 3a). Numbers in circles – xenolith locations discussed in text. (b) Composite model for thermal LAB, based on the TC1 model for the continental part (Artemieva, 2006) and half- space cooling (HSC) model for the oceanic part. TC1 is constrained by heat flux measurements. HSC is constrained by the age of ocean floor. The shelves arenot covered by the TC1 and HSC models, and a constant value of 120 km is assigned here for the lithosphere thickness.

In the eastern part of the Baltic Shield (Kola-Karelia), the litho- with a regional P-wave tomography study which images the LAB beneath sphere thins to ca. 120 km and locally to less than 100 km in the rift Southern Norway and Denmark at 60–70 km depth (Plomerova and system of the White Sea, whereas in the Devonian Kola Alkaline Babuska, 2010). The anomaly is also resolved in global (Fig. 8a) and re- Province the lithosphere thickness is 160–190 km and ca. 160 km thick gional tomography models (cf Maupin et al., 2013; Hejrani et al., 2017). in the Tersk and the Arkhangelsk kimberlite provinces, respectively. In the Russian Platform, thick lithosphere roots (> 250 km) are The latter estimate is somewhat lower than xenolith thermobarometry determined for parts of the Archean Volga-Uralia subcraton (Fig. 7a), estimates from the Devonian Grib kimberlite pipe in the Arkhangelsk where the fastest Vs > 4.7 km/s is observed at 250–320 km depth in region (Afanasiev et al., 2013)(Fig. 9), although the results correspond the upper mantle (Lebedev et al., 2009). While seismic tomography to different time intervals (the Devonian and present). Low seismic shows similar values also for the Archean Ukrainian Shield–Vitebsk velocities at 100–150 km depth in the lithosphere mantle of the Kola block in the western part of the East European craton (Fig. 8), the Alkaline Province have been interpreted as evidence for a strong me- present results indicate a thinner lithosphere of 220–240 km, in tasomatic reworking in the Proterozoic and Paleozoic (Silvennoinen agreement with the lithosphere thermal thickness constrained by sur- et al., 2016). face heat flux (Artemieva, 2006). The Ukrainian Shield has a highly Beneath the Norwegian mountains the lithosphere is less than 140 km heterogeneous lithosphere thickness, which reflects its block structure thick and it thins to ca. 80–100 km in Southern Norway, in agreement and the LAB depth ranges from ca. 220 km to ca. 180 km, decreasing to

Fig. 8. Surface wave tomography models for the lithosphere structure in Europe. (a) Lithosphere thickness defined by a 2% seismic velocity perturbation (model of Shapiro and Ritzwoller, 2002). (b) Shear wave velocity anomalies at a 150 km depth (model of Schaeffer and Lebedev, 2013).

461 I.M. Artemieva Earth-Science Reviews 188 (2019) 454–468

The present results confirm an earlier geological interpretation onthe cratonic basement of the eastern Barents Sea (Nikishin et al., 1996). The lithosphere thickness ranges from ca. 130 km at the eastern Barents Sea basin to ca. 230 km at the offshore extension of the Timanides and the Kola province. The continental lithosphere of the western Barents Sea, that was significantly rifted during the opening of the North Atlantic ocean (Fig. 1), is typically 120–150 km thick and thins to ca. 100 km towards Svalbard. The contrast in the lithosphere structure of the western and eastern Barents Sea has been noted earlier (e.g. Klitzke t al., 2016), but the amplitude of the transition between a thick and thin lithosphere appears to be underestimated in the previous studies. The present study indicates that the boundary between the western and eastern Barents Sea at ca. 34°E is marked by a sharp change in the lithosphere thickness by ca. 50–70 km.

4.4. Phanerozoic Europe

Four different types of lithosphere structure can be recognized in Phanerozoic Europe: the collisional orogens with a thick (> 200 km) lithosphere, the Gondwana massifs with a 160–180 km thick lithosphere, regions with Cenozoic magmatism where the lithosphere is thinned to < 60–80 km, and the rest of the Variscan Europe with a 100–120 km thick lithosphere. These geodynamic types are briefly described below. In regions with a dense seismic data coverage of the Moho depth, such as the Alps, the present results agree with seismic models for the lithosphere thickness (Blundell et al., 1992; Lippitsch et al., 2003). Large lithosphere thickness in the collisional orogens (Alps, Pyrenees, Fig. 9. Typical continental geotherms (based on Pollack and Chapman, 1977). Carpathians, Dinarides) implies the presence of a subducting litho- Color code as in Fig. 7a, 10a. Symbols – xenolith P-T arrays from several lo- sphere plate, in agreement with seismic tomography models which cations in Europe (marked in Fig. 7a). Xenolith P-T arrays are based on data of image subducting slabs beneath the European orogens and a mosaic of Kukkonen and Peltonen (1999), Lehtonen et al. (2004), Malkovets et al. (2003), high- and low-velocity bodies in the upper mantle of southern Europe Werling and Altherr (1997). (Piromallo and Morelli, 2003; Zhu et al., 2012; Palomeras et al., 2017; Hetenyi et al., 2018; Kastle et al., 2018). The present results indicate 140–150 km in the Azov block in the south. The southern Uralides have that the depth extent of the slabs is different: 250–260 km km in the SW a thick lithosphere (200–220 km), in agreement with seismic models Alps and ca. 200–220 km beneath the Pyrenees and the Carpathians. which indicate the presence of a 175–200 km thick seismic lid (Knapp The Vrancea seismic zone at the SE bend of the Carpathian orogen has a et al., 1996; Ryberg et al., 1996). The Proterozoic rifts and aulacogens thin lithosphere (60–80 km), therefore supporting geodynamic models of the Russian Platform slightly differ from the surrounding mantle and for lithosphere delamination and slab detachment (Wortel and have lithosphere thickness of ca. 160–180 km. Spakman, 2000; Lorinczi and Houseman, 2009; Koulakov et al., 2010), Surprisingly, the huge Devonian Dnieper–Donets rift in the southern as is also reflected in depth distribution of local seismicity (e.g. part of the Russian platform is not seen in the lithosphere thickness Cloetingh et al., 2007; Ren et al., 2012). In the Dinarides, the calculated calculated from thermal isostasy, nor in the upper mantle velocity lithosphere thermal thickness of 200–220 km is significantly larger than structure (Lebedev et al., 2009). A localized anomaly in its central part estimates based on seismic S-receiver function analysis (ca. 160 km) with a thinned lithosphere (130–150 km) extends both to the Ukrainian (Belinića et al., 2018). Therefore the results contribute to the on-going Shield and to the Archean Voronezh Massif, and in the Ukrainian Shield debate if seismic RFs image the LAB (Rychert and Shearer, 2009) or a it corresponds to the Korsun pluton. discontinuity with velocity contrast stronger than at the LAB, such as The strongest anomaly with a thin (75–100 km) and hot lithosphere the enigmatic mid-lithosphere discontinuity (Thybo and Perchuc, 1997; in the Peri-Caspian Basin is apparently associated with the Devonian Thybo, 2006; Fischer et al., 2010; Karato et al., 2015). Central rift (Fig. 7a). This region also has a strong positive Vs anomaly The results for the region south of the Black Sea-Caspian Sea at a 100 km depth (Lebedev et al., 2009). Note that both the Dnieper- (Anatolia, the Caucasus, the Caspian Basin, the Turanian plate, and the Donetsk rift and the Peri-Caspian Basin have > 20 km of sediments, the Zagros orogen) should be treated with caution due to the very limited crystalline basement is thinned to < 20 km (cf. Artemieva and Thybo, information available for the Moho depth in these regions (Fig. 6b). In 2013), and therefore the average crustal density is significantly smaller particular, it is challenging to provide any constraints on the litho- than in normal platform crust (Fig. 3b). The proposed method provides sphere structure in the Lesser Caucasus, Anatolia and the northern Ir- reliable constraints on the lithosphere thickness even in these regions, anian Plateau. Furthermore, these regions are not in the isostatic therefore validating the assumption that much of crustal density het- equilibrium (Fig. 1) and, strictly speaking, the proposed thermal iso- erogeneity is compensated by Moho depth variations. stasy method is non-applicable there. The Gondwana massifs of western Europe (Armorican, Ardennes, Iberian, Bohemian, London-Brabant, and French Massif Central) are the 4.3. The shelf terranes of Proterozoic to early Paleozoic ages that have been deformed and reworked during the Variscan orogeny (cf. Artemieva et al., 2006). The off-shore extent of the cratonic lithosphere into the Barents Sea Several of these terranes are clearly seen as blocks with a relatively shelf is debated (e.g. Ritzmann and Faleide, 2009; Klitzke et al., 2016). thick (160–180 km) and cold lithosphere (Fig. 7a). This includes the Seismic tomography models show high-velocity lithosphere beneath the Armorican Massif where fast P-wave velocity anomaly is observed shelf (Levshin et al., 2007; Lebedev et al., 2018), but differ in details. down to a 150 km depth (Judenherc et al. 2002), and the London-

462 I.M. Artemieva Earth-Science Reviews 188 (2019) 454–468

Fig. 10. Geothermal heat flux. (a) Present study, calculated from the lithosphere thermal thickness (Fig. 7a). (b) Measured heat flux based on the open-access database of the International Heat Flow Commission.

Brabant and the Bohemian Massifs where the lithosphere is ca. lithosphere (upper mantle) material down to ca. 200 km depth. 180–200 km and ca. 120–140 km thick, correspondingly. Both massifs However, the information on the Moho depth in the Black Sea is in- exhibit low seismicity and have mechanically strong, old, cold litho- sufficient for reliable constraints. Note that, in agreement withthe sphere (Cloetingh et al., 2007). The LAB depth beneath the Bohemian present results, seismic tomography models also indicate the presence Massif from P-wave tomography is ca. 120 km (Plomerova and of a cold and thick lithosphere in the Black Sea region and, possibly, in Babuska, 2010). The lithosphere of the Iberian Massif is relatively thin, the Bay of Biscay (Fig. 8). 80–120 km, in agreement with recent tomography studies that indicate the presence of a low shear-velocity material below a 70–100 km depth 5. Thermal structure of the European mantle and geothermal heat (Palomeras et al., 2017). flux The European Cenozoic Rift System that extends from southern France to the North Sea is marked by a thin and hot lithosphere Knowledge of the lithosphere thermal thickness allows for de- (< 80 km) of the Massif Central, Rhenish Massif and the Rhine Graben termining regional geotherms and for predicting geothermal heat flux. (Fig. 7a) and the results are in a good agreement with xenolith geo- The latter possibility is important because it allows for comparing the thermobarometry for the Massif Central (Fig. 9)(Werling and Altherr, model predictions with heat flux measurements. A few comments are 1997). The presence of a thin (60–100 km), hot and mechanically weak worth making here. First, the quality of heat flux measurements is lithosphere in these regions, that have been affected by the Cenozoic variable and at least half of the borehole measurements in Europe are magmatism, has long been reported in regional seismic (Babuska et al., shallow. It means, in particular, that in many cases paleoclimate 2002; Granet et al., 1995) and thermal studies (Cermak and Bodri, correction is required, and in the Baltic Shield it amounts up to ca. 1995; Sobolev et al., 1996). 25% of the measured values (Artemieva, 2006; Majorowicz and The Pannonian Basin whose formation has been affected by roll- Wybraniec, 2011). Second, strong heterogeneity in radioactive heat back of the Carpathian slab has an extremely thin lithosphere production, in particularly in sedimentary strata and the upper crust, (< 60 km). This conclusion is in agreement with numerous seismic produces short-wavelength anomalies in heat flux, which may not be tomography, magnetotelluric, gravity and thermal studies that in- associated with variations in the lithosphere thermal thickness. Third, dicate a lithosphere thickness of 45–80 km beneath the Pannonian in regions of active deformation and Cenozoic magmatism a sig- Basin (Adam, 1996; Tari et al., 1999; Horvath et al., 2006). Similar nificant part of the measured heat flux may be advective andasso- small values of the lithosphere thickness are calculated here for the Po ciated with active faults. Likewise, hydrothermal circulation in ocean Basin. sediments is known to have a strong effect on heat flux measurements in oceans. 4.5. Oceanic lithosphere A comparison of the predicted and measured heat flux demonstrates that the thermal isostasy method provides results that are in a good The results indicate a strongly heterogeneous lithosphere of the agreement with observations (Figs. 10, 11). A comparison with the map North Atlantic ocean, with a possible presence of continental fragments, of data coverage for the crustal structure (Fig. 6b) indicates that a major in particular between the Aegir paleo-spreading ridge and the reason for the discrepancy is the poorly known Moho depth. Regions Norwegian margin (Fig. 7a). This anomaly appears to be connected to where the predicted and measured heat flux values disagree by >20 the Archean Lofoten block in northern Norway. mW/m2 include mostly the Mediterranean region (Fig. 11a) with a A large amplitude anomaly is associated with the Bay of Biscay, complex mosaic of subduction systems, active volcanic systems, and suggesting the presence of a thick cold lithosphere, similar to the cra- poor knowledge on the crustal structure. tonic. The result is apparently in contradiction with strong positive The pattern of the predicted heat flux matches the pattern of the Bouguer anomalies in the region. A similar pattern is observed for the lithosphere thickness variations and therefore it is not described here Black Sea, where the results question the presence of oceanic litho- in detail, except for some regions. As expected, the Archean blocks sphere in this basin, but rather indicate the presence of a cold (central Finland, the Vitebsk Massif, and the Ukrainian Shield) have

463 I.M. Artemieva Earth-Science Reviews 188 (2019) 454–468

Fig. 11. Comparison between the predicted (Fig. 10a) and measured (Fig. 10b) surface heat flux. (a) Heat flux difference based on interpolation (interpolation radius 150 km). Gray shading - regions with no heat flux measurements. For most of the area, predicted and measured heat flux agree within±10mW/m2. Major differences are in tectonically active regions with a highly heterogeneous heat flux and a likely advective component to measured values. (b) Statistical analysisof predicted and measured heat flux difference. a low heat flux, < 40 mW/m2, in agreement with measurements therefore providing independent data to verify the model predictions. A which in many boreholes provide anomalous values of less than large number of regional seismic models allow for further comparison 25–30 mW/m2. of the model predictions on the lithosphere thickness with seismic The present results predict low lithosphere temperatures and low constraints. The results are in a good agreement with seismic models heat flux in the southern Urals. This region is well known foranex- and borehole heat flux measurements and allow to make the following tremely low surface heat flux (~25 mW/m2 as compared to generally conclusions. 40–50 mW/m2 in the cratonic Russian Platform), and the origin of this thermal anomaly has long been debated (Kukkonen et al., 1997). The 1. Thick, cold and mechanically strong lithosphere is present in most present results suggest a deep lithosphere origin for the anomaly, rather of the East European craton. The largest thickness (250–300 km) is than a shallow source related to a low crustal heat production, hydro- in the Proterozoic central Finland and in some Archean blocks of thermal circulation or paleoclimate effects. the Russian platform. Most of the East European craton has a ca. The transition from the cratonic to the non-cratonic lithosphere is 160–220 km thick thermal lithosphere, and similar lithosphere marked by a change in geothermal heat flux from typical cratonic va- structure is typical of the southern Urals. lues of 40–50 mW/m2 to 80–90 mW/m2 in most of Phanerozoic Europe. 2. Southern Norway has an unusually thin lithosphere (70–100 km). Areas with a high heat flux and non-steady-state thermal regime in- This region has experienced fast recent uplift, which may have been clude primarily the regions of Cenozoic magmatism in the Massif caused or facilitated by removal of the basal part of the cratonic Central and the Rhine graben, and also the Pannonian Basin, where lithosphere mantle. both the predicted and measured heat flux (up to 130 mW/m2) show a 3. Proterozoic and Paleozoic rifts of the East European craton have a significant departure from the average surface heat flux inthe typical cratonic lithosphere thermal structure. The exceptions are Phanerozoic Europe. the North Caspian Basin and the rifted margin of the White Sea, In the ocean domain, the model predicts an unusually low heat flux where the lithosphere is locally thinned to < 100 km. in the Bay of Biscay, the Black Sea, and off-shore Norway. These model 4. The Barents Sea is underlain by the cratonic lithosphere, which is predictions are in agreement with atypically low heat flux values thick (ca. 200 km on average) in its eastern part, and thin (ca. measured in these off-shore regions. 100–140 km) in the western rifted part. The transition between the two domains is sharp and possibly step-wise. 6. Conclusions 5. The craton-non-craton transition in the lithosphere mantle deviates from its surface expression at the Trans-European Suture Zone by This study presents a new method of thermal isostasy to model the up to ca. 200 km laterally. This suggests that the edge of the craton lithosphere thermal thickness, lithosphere geotherms, and geothermal has been significantly reworked and deformed. In the Baltic Shield, heat flux from two parameters: seismically constrained Moho depth and the western edge of the thick and cold cratonic lithosphere ap- topography (bathymetry). The method is applied to the European proximately corresponds to the western margin of the Proterozoic mantle, where a dense network of seismic reflection/refraction profiles Trans-Scandinavian Igneous Belt. provides detailed information on the Moho depth for most of the region. 6. The Paleoproterozoic – early Paleozoic Gondwana massifs of wes- The region is also densely covered by borehole heat flux measurements, tern Europe are clearly recognized by a thick (120–180 km) and

464 I.M. Artemieva Earth-Science Reviews 188 (2019) 454–468

cold lithosphere, except for the Iberian Massif where thermal li- connection to the Archean Lofoten block. thosphere is ca. 100 km thick, similar to most of the Variscan 10. Strong anomalies in the Bay of Biscay and the Black Sea may be Europe. caused by a cold material in the upper mantle down to a 200 km 7. The Cenozoic orogens have lithosphere extending down to ca. depth. 200–250 km depth, with the deepest slab under the Alps. The Vrancea block of the Carpathians orogen has a thin (< 80 km) lithosphere, in support of models of lithosphere delamination. Acknowledgements 8. The European Cenozoic Rift System is marked by a thin and hot lithosphere (< 60–80 km). The presence of an extremely thin li- Constructive comments from A.Watts and anonymous referees are thosphere (50–60 km) beneath the Pannonian and Po Basins may gratefully acknowledged. Research is supported by grant DFF-1323- indicate the onset of oceanization. 00053 of the Danish Fund for Independent Research (DFF-FNU). 9. Strong lithosphere heterogeneity of the North Atlantic ocean sug- Digital model for the lithosphere thermal thickness (Fig. 7a) is gests the presence of continental lithosphere fragments between the available as electronic supplement from Mendeley depository, or Aegir paleo-spreading and the Norwegian margin, with a possible from the author.

Appendix A. Sensitivity analysis

The set of assumptions and the often unknown uncertainty in the input parameters makes it practically impossible to evaluate the accuracy of model predictions. I therefore run a series of tests, varying the input parameters within a possible uncertainty range to estimate the uncertainty of modeling.

(1) Uncertainties in density values for the bulk crust and bulk lithosphere mantle ρo (eq. 4) and the asthenosphere mantle ρm (eq. 3) may affect the 3 3 constant in eqs. (6) and (9). In particular, a change of ρm from 3.40 t/m to 3.35 t/m will change the constant from 94,683 to 93,212, and therefore the final result will change by 1.5%. A change by 30.10 t/m in average SPT density of the bulk crust from the assumed 2.85 t/m3 has negligible effect on the bulk lithosphere density (eq. 4). Similarly, a change of average SPT density of the lithosphere mantle from 3.35 t/m3 to 3.30 t/m3 will change average SPT lithosphere density by ca. 1.2% for a 150 km thick lithosphere (eq. 4), leading to a < 1% change in in situ lithosphere density (eq. 5) and the calculated ΔT (eq. 6). (2) Uncertainty in the Moho depth (typically 1–2 km for seismic reflection/refraction studies) contributes to the final uncertainty through theMoho depth-versus-topography correlation, where its effect is statistically insignificant in case of large regional data sets for both parameters (seealso the next point). The effect of the Moho depth uncertainty on the calculated ΔT (eq. 6) is also negligible. Therefore, except for specific regions with limited or controversial data on the Moho depth, which affects the Moho depth-versus-topography correlation, the uncertainty in theMoho depth has little effect on the model results. It is worth noting that crustal models where the Moho depth is constrained by gravity data(including parts of global crustal models CRUST1.0 and 2.0 where gravity constraints are used in regions not covered by seismic profiles) are not suitable for the proposed thermal isostasy method. (3) Uncertainty in the best-fit correlation between the Moho depth and topography affects the slope and the intersect values of the best-fit line.Both parameters are controlled by regional tectonics. The slope depends on the density contrast across the Moho and therefore it has a narrow range of possible values, 4–7 for bulk crustal density of 2.70–2.90 t/m3 (Fig. A1). The intercept reflects the average crustal thickness in the region of interest and for typical continental crust it may differ by 5–10 km.

I further illustrate this effect for western Europe where the data are best fit by an intercept of 33.4 km at zero topography (see section 3),while global continental statistics suggests that the intercept is ca. 38 km (Christensen and Mooney, 1995). Different values for the intercept value lead to uncertainty in the calculated values of anomalous topography by causing its systematic shift by a constant value, if the slope of the Moho depth- topography best-fit is the same. In case of a different slope as it is the case for the European continent, the shift is proportional totopography(Fig. A2a) and the analysis may require to split data into subdomains with different relationships between the Moho depth and topography. This will improve the model constraints and will remove a correlation with tectonic settings (Fig. A2a). Due to a change in A, the calculated temperature anomaly and the LAB depth also become systematically shifted when a different Moho depth – topography relationship is used. However, in this case reference points (locations with A = 0 km) also become different (Fig. A2a) and Ref therefore zLAB values (the LAB depth at reference points) are also different (eq. 9). Due to this complex link between the Moho depth – topography relationship and the lithosphere thickness at reference points, the results are robust. In Europe, when the same relationship for the Moho depth versus topography is used for the entire continent (discussed below), the LAB depth changes by ca. 25 km on average (Fig. A2b). As mentioned above, the difference can be reduced if tectonic regions with large deviations are modeled separately, based on regional Mohodepth – topography relationships. Importantly, the model results can be verified by independent data, such as heat flux measurements. This allowsto adjust the best-fit line for the Moho depth – topography relationship to match heat flux observations. This procedure is not used inthepresent study. 4) Uncertainty in the LAB depth at reference points affects the calculated values of the lithosphere thermal thickness (eq. 9) and the predicted heat flux. Importantly, the LAB depth at reference points usually is not the unknown parameter, since it can be determined from independentdata (seismic, thermal, etc). For regions with lithosphere thermal thickness > 100 km (most of the continents) an increase in the reference LAB depth by 50 km increases the calculated lithosphere thickness by ca. 20 km, but the effect is strongly non-linear and it is significantly smaller for regions witha thin lithosphere (Artemieva, 2018). The overall uncertainty in the modeled lithosphere thickness is < 50 km (typically 10–30 km), which is comparable or even better than in seismic tomography. This uncertainty propagates into the uncertainty of the predicted heat flux (10–15% and 5–10% for a 100 km and 200 kmthick lithosphere, correspondingly).

465 I.M. Artemieva Earth-Science Reviews 188 (2019) 454–468

Fig. A1. Correlation between bulk in situ crustal density (5 cases are shown), bulk in situ density of the lithospheric mantle, and the slope of the Moho depth versus topography correlation based on the Airy-type crustal isostasy.

Fig. A2. Sensitivity analysis which shows how the parameters in the correlation between the Moho depth and topography affect the results for the western Europe. Two Moho-topography correlations are used: regional for the western Europe and global continental. (a) Difference in calculated anomalous topography A for regional and global continental Moho-topography correlations. (b) Difference in the lithosphere thickness calculated based on (a) regional and globalMoho- topography correlations. Diamonds and crosses in (a) show some locations with A = 0 km for global and regional Moho-topography correlations, respectively. Ref Calculation of the lithosphere thickness requires an assumption on the LAB depth zLAB in regions with A = 0 km, which is set at 200 km for the global Moho- topography correlation and at 150 km for the regional correlation. The difference for the predicted LAB depth based on regional and global continental Moho- topography correlations is on average < 30 km, which is comparable or better than the depth resolution of seismic tomography models. A comparison between the predicted and measured heat flux allows for further tuning of the parameters.

References Belinića, T., Stipčevića, J., Živčić, M., AlpArrayWorking Group, 2018. Lithospheric thickness under the Dinarides. Earth Planet. Sci. Letters 484, 229–240. Blundell, D., Freeman, R., Mueller, S. (Eds.), 1992. A continent revealed. The European Adam, A., 1996. Regional magnetotelluric (MT) anisotropy in the Pannonian Basin Geotraverse. Cambridge University Press. (Hungary). Acta Geodaetica of Geophysica Hungavicae 31, 191–216. Boyd, F.R., 1989. Compositional distinction between oceanic and cratonic lithosphere. Afanasiev, V.P., Ashchepkov, I.V., Verzhak, V.V., O'Brien, H., Palessky, S.V., 2013. P-T Earth Planet. Sci. Lett. 96, 15–26. conditions and trace element variations of picroilmenites and pyropes from the Bruneton, M., Pedersen, H., Farra, V., et al., 2004. Complex lithospheric 651 structure Arkhangelsk region. Journal of Asian Earth Sciences 45–63. under the Central Baltic Shield from surface wave tomography. J. Geophys. Res. 109 Artemieva, I.M., 2006. Global 1°×1° thermal model TC1 for the continental lithosphere: (B10), 652. Implications for lithosphere secular evolution. Tectonophysics 416, 245–277. Carlson, R.L., Herrick, C.N., 1990. Densities and porosities in the oceanic-crust and their Artemieva, I.M., 2009. The continental lithosphere: Reconciling thermal, seismic, and variations with depth and age. J. Geophys. Res. 95, 9153–9170. petrologic data. Lithos 109, 23–46. Carlson, R.W., Pearson, G., James, D.E., 2005. Physical, chemical, and chronological Artemieva, I.M., 2011. The lithosphere: An interdisciplinary approach. 794 Cambridge characteristics of continental mantle. Rev. Geophys. 43, RG1001. Univ. Press. Cermak, V., Bodri, L., 1995. Three-dimensional deep temperature modeling along the Artemieva, I.M., 2018. Lithosphere thermal thickness and geothermal heat flux in European Geotraverse. Tectonophysics 244, 1–12. Greenland from a new thermal isostasy method. Earth Science Reviews. https://doi. Cermak, V., Rybach, L., 1979. Terrestrial heat flow in Europe. 330 Springer. org/10.1016/j.earscirev.2018.10.015. Cherepanova, Y., Artemieva, I.M., 2015. Density heterogeneity of cratonic lithospheric Artemieva, I.M., Mooney, W.D., 2001. Thermal thickness and evolution of Precambrian mantle: A case study of the Siberian craton. Gondwana Research 28, 1344–1360. lithosphere: a global study. J. Geophys. Res. 106, 16,387–16,414. Christensen, N.I., Mooney, W.D., 1995. Seismic velocity structure and composition of the Artemieva, I.M., Thybo, H., 2013. EUNAseis: a seismic model for Moho and crustal continental crust: A global view. J. Geophys. Res. 100, 9761–9788. structure in Europe, Greenland, and the North Atlantic region. Tectonophysics 609, Cloetingh, S.A.P.L., Burov, E., Matenco, L., et al., 2004. Thermo-mechanical controls on 97–153. the mode of continental collision in the SE Carpathians (Romania). Earth Planet. Sci. Artemieva, I.M., Vinnik, L.P., 2016. Density structure of the cratonic mantle in southern Lett. 218, 57–76. Africa: 2. Correlations with kimberlite distribution, seismic velocities and Moho Cloetingh, S.A.P.L., Ziegler, P.A., Bogaard, P.J.F., et al., 2007. TOPO-EUROPE: The sharpness. Gondwana Res. 36, 14–27. geoscience of coupled deep Earth-surface processes. Global Planet. Change 58, Artemieva, I.M., Thybo, H., Kaban, M.K., 2006. Deep Europe today: geophysical synthesis 1–118. of the upper mantle structure and lithospheric processes over 3.5 Ga. Geological Deschamps, F., Trampert, J., Snieder, R., 2002. Anomalies of temperature and iron in the Society, London, Memoirs 32, 11–41. uppermost mantle inferred from gravity data and tomographic models. Phys. Earth Artyushkov, E.V., 1973. Stresses in the lithosphere caused by crustal thickness in- Planet. Int. 129, 245–264. homogeneities. J. Geophys. Res. 78, 7675–7708. Eeken, T., Goes, S., Pedersen, H.A., Arndt, N.T., Bouilhol, P., 2018. Seismic evidence for Babuška, V., Montagner, J.P., Plomerova, J., et al., 1998. Age-dependent large-scale depth-dependent metasomatism in cratons. Earth Planet. Sci. Lett. 491, 148–159. fabric of the mantle lithosphere as derived from surface-wave velocity anisotropy. Fischer, K.M., Ford, H.A., Abt, D.L., Rychert, C.A., 2010. The lithosphere–asthenosphere Pure and Appl. Geophys. 151, 257–280. boundary. Ann. Rev. Earth Planet. Sci. 38, 551–575. Babuska, V., Plomerova, J., Vecsey, L., Granet, M., Achauer, U., 2002. Seismic anisotropy Fleitout, L., Froidevaux, C., 1982. Tectonics and topography for a lithosphere containing of the French Massif Central and predisposition of Cenozoic rifting and volcanism by density heterogeneities. Tectonics 1, 21–56. Variscan suture hidden in the mantle lithosphere. Tectonics 21 (4). https://doi.org/ Forte, A.M., Mitrovica, J.X., 1996. New inferences of mantle viscosity from joint inversion 10.1029/2001TC901035. of long-wavelength mantle convection and post-glacial rebound data. Geophys. Res. Beaumont, C., 1981. Foreland basins. Geophysical J. Royal Astronom. Soc. 65, 291–329. Lett. 23, 1147–1150.

466 I.M. Artemieva Earth-Science Reviews 188 (2019) 454–468

Forte, A.M., Perry, A.C., 2000. Seismic-geodynamic evidence for a chemically depleted regional paleoclimatic correction application. Intern. J. Earth Sciences 100 (4), continental tectosphere. Science 290, 1940–1944. 881–887. Fowler, C.M.R., 2005. The solid Earth; An introduction to global geophysics, 2nd edition. Malkovets, V.G., Taylor, L.A., Griffin, W., et al., 2003. Cratonic conditions beneath Cambridge Univ. Press. Arkhangelsk, Russia: garnet peridotites from the Grib Kimberlite. In: Proc. 8th Intern. Gaul, O. F., Griffin, W. L., O'Reilly, S.Y., and Pearson, N. J., 2000. Mapping olivine Kimberlite Conf., Victoria, BC, Canada, June 2003, FLA–0220. composition in the lithospheric mantle. Earth Planet. Sci. Lett., 182, 223–235. Maupin, V., Agostini, A., Artemieva, I., et al., 2013. The deep structure of the Scandes and Geissler, W.H., Sodoudi, F., Kind, R., 2010. Thickness of the central and eastern European its relation to tectonic history and present-day topography. Tectonophysics 602, lithosphere as seen by S receiver functions. Geophys. J. Intern. 181, 604–634. 15–37. Goes, S., Govers, R., Vacher, P., 2000. Shallow mantle temperatures under Europe from P Nikishin, A.M., Ziegler, P.A., Stephenson, R.A., et al., 1996. Late Precambrian to Triassic and S wave tomography. J. Geophys. Res. 105, 11153–11169. history of East European Craton: dynamics of sedimentary basin evolution. Granet, M., Wilson, M., Achauer, U., 1995. Imaging a mantle plume beneath the French Tectonophysics 268, 23–63. Massif Central. Earth Planet. Sci. Letters 136, 281–296. Norden, B., Förster, A., Balling, N., 2008. Heat Flow and lithospheric thermal regime in Griffin, W.L., O'Reilly, S.Y., 2007. The earliest subcontinental lithospheric mantle. In:van the Northeast German Basin. Tectonophysics 460, 215–229. Kranendonk, M.J., Smithies, R.H., Bennett, V.C. (Eds.), Developments in Precambrian Palomeras, I., Villasenor, A., Thurner, S., et al., 2017. Lithospheric structure of Iberia and Geology, Earth's Oldest Rocks. 15. Elsevier, pp. 1013–1035. Morocco using finite-frequency Rayleigh wave tomography from earthquakes and Hejrani, B., Balling, N., Jacobsen, B.H., et al., 2017. Upper-mantle velocities below the seismic ambient noise. Geochemistry Geophysics Geosystems 18, 1824–1840. Scandinavian Mountains from P- and S-wave traveltime tomography. Geophys. J. Pavlis, N.K., Holmes, S.A., Kenyon, S.C., et al., 2012. The development and evaluation of Intern. 208, 177–192. the Earth Gravitational Model 2008 (EGM2008). J. Geophys. Res. 117, B04406. Hetenyi, G., Molinari, I., Clinton, J., et al., 2018. The AlpArray Seismic Network: A Large- Pedersen, H.A., Debayle, E., Maupin, V., the POLENET/LAPNET Working Group, 2013. Scale European Experiment to Image the Alpine Orogen. Surveys Geophys. 39, Strong lateral variations of lithospheric mantle beneath cratons – Example from the 1009–1033. Baltic Shield. Earth Planet. Sci. Lett. 383, 164–172. Horvath, F., Bada, G., Szafian, P., et al., 2006. Formation and deformation ofthe Piromallo, C., Morelli, A., 2003. P wave tomography of the mantle under the Pannonian Basin: Constraints from observational data. Geol. Soc. London Memoirs 32 Alpine–Mediterranean area. J. Geophys. Res. 108, B22065. (1), 191–206. Plomerova, J., Babuska, V., 2010. Long memory of mantle lithosphere fabric — European Jones, A.G., Plomerova, J., Korja, T., et al., 2010. Europe from the bottom up: A statistical LAB constrained from seismic anisotropy. Lithos. https://doi.org/10.1016/j.lithos. examination of the central and northern European lithosphere–asthenosphere 2010.01.008. boundary from comparing seismological and electromagnetic observations. Lithos Plomerova, J., Kouba, D., Babuska, V., 2002. Mapping the lithosphere-asthenosphere 120, 14–29. boundary through changes in surface-wave anisotropy. Tectonophysics 358, Jordan, T.H., 1981. Continents as a chemical boundary layer. Phil. Trans. Roy. Soc. 175–185. London, Ser. A. 301, 359–373. Plomerova, J., Vecsey, L., Babuska, V., LAPNET Working Group, 2011. Domains of Jordan, T.H., 1988. Structure and formation of the continental tectosphere. J. Petrol. 29, Archean mantle lithosphere deciphered by seismic anisotropy – inferences from the 11–37. LAPNET array in northern Fennoscandia. Solid Earth 2, 303–313. Judenherc, S., Granet, M., Brun, J.-P., et al., 2002. Images of lithospheric heterogeneities Pollack, H.N., Chapman, D.S., 1977. On the regional variation of heat flow, geotherms in the Armorican segment of the Hercynian Range in France. Tectonophysics 358, and the thickness of the lithosphere. Tectonophys. 38, 279–296. 121–134. Pollack, H.N., Hurter, S.J., Johnson, J.R., 1993. Heat flow from the Earth's interior: Karato, S.-I., Olugboji, T., Park, J., 2015. Mechanisms and geologic significance of the analysis of the global data set. Rev. Geophys. 31, 267–280. mid-lithosphere discontinuity in the continents. Nature Geoscience 8, 509–515. Poudjom Djomani, Y.H., O’Reilly, S.Y., Griffin, W.L., Morgan, P., 2001. The density Kastle, E.D., El-Sharkawy, A., Boschi, L., et al., 2018. Surface Wave Tomography of the structure of subcontinental lithosphere through time. Earth Planet. Sci. Lett. 184, Alps Using Ambient-Noise and Earthquake Phase Velocity Measurements. J. Geophys. 605–621. Res. 123, 1770–1792. Praus, O., Pecova, J., Petr, V., Babuska, V., Plomerova, J., 1990. Magnetotelluric and Klitzke, P., Sippel, J., Faleide, J.I., Scheck-Wenderoth, M., 2016. A 3D gravity and thermal seismological determination of the lithosphere–asthenosphere transition in central- model for the Barents Sea and Kara Sea. Tectonophysics 684, 131–147. Europe. Physics Earth Planet. Inter. 60 (1–3), 212–228. Knapp, J.H., Steer, D.N., Brown, L.D., et al., 1996. Lithosphere-scale seismic image of the Ren, Y., Stuart, G.W., Houseman, G.A., et al., 2012. Upper mantle structures beneath the Southern Urals from explosion-source reflection profiling. Science 274, 226–228. Carpathian–Pannonian region: Implications for the geodynamics of continental col- Korja, T., 2007. How is the European lithosphere imaged by magnetotellurics? Surv. lision. Earth Planet. Sci. Lett. 349/350, 139–152. Geophys. 28, 239–272. Ritzmann, O., Faleide, J.-I., 2009. The crust and mantle lithosphere in the Barents Sea/ Koulakov, I., Zaharia, B., Enescu, B., et al., 2010. Delamination or slab detachment be- Kara Sea region. Tectonophysics 470, 89–104. neath Vrancea? New arguments from local earthquake tomography. Geochem. Romanowicz, B., 2009. The Thickness of Tectonic Plates. Science 324 (5926), 474–476. Geophys. Geosystems 11 (3). https://doi.org/10.1029/2009GC002811. Ryberg, T., Wenzel, F., Mechie, J., Egorkin, A., Fuchs, K., Solodilov, L., 1996. Two-di- Kukkonen, I.T., Peltonen, P., 1999. Xenolith-controlled geotherm for the central mensional velocity structure beneath Northern Eurasia derived from the super long- Fennoscandian Shield: implications for lithosphere-asthenosphere relations. range seismic profile Quartz. Bull. Seismol. Soc. America 86, 857–867. Tectonophysics 304, 301–315. Rychert, C.A., Shearer, P.M., 2009. A global view of the lithosphere-asthenosphere Kukkonen, I.T., Golovanova, I.V., Khachay, Yu.V., et al., 1997. Low geothermal heat flow boundary. Science 324 (5926), 495–498. of the Urals fold belt – implications of low heat production, fluid circulation or pa- Schaeffer, A., Lebedev, S., 2013. Global shear speed structure of the upper mantleand leoclimate? Tectonophysics 276, 63–85. transition zone. Geophys. J. Int. 194, 417–449. Kukkonen, I.T., Kinnunen, K.A., Peltonen, P., 2003. Mantle xenoliths and thick litho- Schaeffer, A.J., Lebedev, S., 2015. Global heterogeneity of the lithosphere and underlying sphere in the Fennoscandian Shield. Physics Chem. of the Earth 28, 349–360. mantle: a seismological appraisal. In: Khan, A., Deschamps, F. (Eds.), The Earth's Lebedev, S., Boonen, J., Trampert, J., 2009. Seismic structure of Precambrian lithosphere: heterogeneous mantle. Springer, pp. 3–34. New constraints from broad-band surface-wave dispersion. Lithos 109, 96–111. Schutt, D.L., Lesher, C.E., 2006. Effects of melt depletion on the density and seismic Lebedev, S., Schaeffer, A.J., Fullea, J., Pease, V., 2018. Seismic tomography of the Arctic velocity of garnet and spinel lherzolite. J. Geophys. Res. 111, B5. https://doi.org/10. region: Inferences for the thermal structure and evolution of the lithosphere. Geol. 1029/2003jb002950. Soc., London Sp. Publ. 460, 419–444. Shapiro, N.M., Ritzwoller, M.H., 2002. Monte–Carlo inversion for a global shear velocity Lee, C.-T.A., 2003. Compositional variation of density and seismic velocities in natural model of the crust and upper mantle. Geophys. J. Int. 151, 1–18. peridotites at STP conditions: Implications for seismic imaging of compositional Silvennoinen, H., Kozlovskaya, E., Kissling, E., 2016. POLENET/LAPNET teleseismic P heterogeneities in the upper mantle. J. Geophys. Res. 108 (B9), 2441. wave travel time tomography model of the upper mantle beneath northern Lee, C.-T.A., 2006. Geochemical/petrologic constraints on the origin of cratonic mantle. Fennoscandia. Solid Earth 7, 425–439. In: Benn, K., Maraschel, J.-C., Condie, K.C. (Eds.), Archean Geodynamics and Simmons, N.A., Forte, A.M., Grand, S.P., 2009. Joint seismic, geodynamic and mineral Environments. American Geophysical Union, pp. 164 Geophysical Monograph Series. physical constraints on three-dimensional mantle heterogeneity: Implications for the Lehtonen, M., O'Brien, H., 2009. Mantle transect of the Karelian Craton from margin to relative importance of thermal versus compositional heterogeneity. Geophys. J. Int. core based on P-T data from garnet and clinopyroxene xenocrysts in kimberlites. Bull. 177, 1284–1304. Geol. Soc. Finland 81, 79–102. Sobolev, S.V., Zeyen, H., Stoll, G., Werling, F., Altherr, R., Fuchs, K., 1996. Upper mantle Lehtonen, M., O'Brien, H. E., Peltonen, P., Johanson, B. S., and Pakkanen, L. 2004. temperatures from teleseismic tomography of French Massif Central including effects Layered mantle at the Karelian Craton margin: P-T of mantle xenocrysts and xenoliths of composition, mineral reactions, anharmonicity, anelasticity and partial melt. Earth from the Kaavi–Kuopio kimberlites, Finland, Lithos, 77, 593–608. Planet. Sci. Lett. 139, 147–164. Levshin, A.L., Schweitzer, J., Weidle, C., Shapiro, N.M., Ritzwoller, M.H., 2007. Surface Soomro, R.A., Weidle, C., Cristiano, L., Lebedev, S., Meier, T., PASSEQ Working Group, wave tomography of the Barents Sea and surrounding regions. Geophys. J. Int. 170, 2016. Phase velocities of Rayleigh and Love waves in central and northern Europe 441–459. from automated, broad-band, interstation measurements. Geophys. J. Intern. 204, Lippitsch, R., Kissling, E., Ansorge, J., 2003. Upper mantle structure beneath the Alpine 517–534. orogen from high-resolution teleseismic tomography. J. Geophys. Res. 108 (B8), Tari, G., Dövényi, P., Dunkl, I., et al., 1999. Lithospheric structure of the Pannonian basin 2376. derived from seismic, gravity and geothermal data. In: Geol. Soc., London, Sp. Publ. Lorinczi, P., Houseman, G.A., 2009. Lithospheric gravitational instability beneath the 156. pp. 215–250. Southeast Carpathians. Tectonophysics 486, 150–161. Thybo, H., 2006. The heterogeneous upper mantle low velocity zone. Tectonophysics Ludwig, W.J., Nafe, J.E., Drake, C.L., 1970. Seismic refraction. In: Maxwell, A.E. (Ed.), 416, 53–79. The Sea, Ideas and Observations on Progress in the Study of the Seas. 4. Wiley- Thybo, H., Perchuc, E., 1997. The seismic 8° discontinuity and partial melting in con- Interscience, New York, pp. 53–84. tinental mantle. Science 275, 1626–1629. Majorowicz, J., Wybraniec, S., 2011. New terrestrial heat flow map of Europe after Vinnik, L.P., Erduran, M., Oreshin, S.I., et al., 2014. Joint inversion of P- and S-receiver

467 I.M. Artemieva Earth-Science Reviews 188 (2019) 454–468

functions and dispersion curves of Rayleigh waves: the results for the Central Massif Central as deduced from geothermobarometry on mantle xenoliths. Anatolian Plateau. Izv. Phys. Solid Earth 50, 622–631. https://doi.org/10.1134/ Tectonophysics 275, 119–141. S106935131404017X. Wortel, M.J.R., Spakman, W., 2000. Subduction and slab detachment in the Vinnik, L.P., Kozlovskaya, E., Oreshin, S., Kosarev, G., Piiponen, K., Silvennoinen, H., Mediterranean–Carpathian region. Science 290, 1910–1917. 2016. The lithosphere, LAB, LVZ and Lehmann discontinuity under central Yang, Y., Ritzwoller, M.H., Levshin, A.L., Shapiro, N.M., 2007. Ambient noise Rayleigh Fennoscandia from receiver functions. Tectonophysics 667, 189–198. wave tomography across. Europe. Geophys. J. Intern. 168, 259–274. Watts, A.B., 2001. Isostasy and Flexure of the Lithosphere. Cambridge Univ. Press. Zhu, H., Bozdag, E., Peter, D., Tromp, J., 2012. Structure of the European upper mantle Werling, F., Altherr, R., 1997. Thermal evolution of the lithosphere beneath the French revealed by adjoint tomography. Nature Geoscience 5, 493–498.

468